Make small improvements re HFS lines

See merge request !41

Previously, a method was defined, LinSv.isSpectralLine(), that
discriminated against all line types and flags known to the code, except
for lines that are either transferred, 't', or are designated as
recombination, 'r', lines.  Then, when the method was needed there was
an additional check on the TransitionProxy of the line in the stack.

Robin wondered if all that was necessary, so upon a little reflection
it became apparent that they were not.  Instead of testing against all
those flags, it is more straightforward to test for the types that the
method does allow for: 't' and 'r'.

But given that only 't' lines with initialized transitions can ever pass
the if-statement in 'save lines pop', and produce meaningful output, the
method becomes superfluous.

It is hereby removed.  The if-statement now checks if the line is type
't' instead of calling that method to test if it is a genuine spectral
line.

The check against the transition being initialized (tha...

Remove tsuite/bugs that crept in when merging ca1_cmesh.

Remove tsuite/bugs dir

See merge request !43

Add commands for each and every of the command's options to sim
limit_caseb_he_den2_temp4.in.

Inform users that if the requested lines are not associated with an
atomic model, the command will report zeros.

The previous implementation of the command rested on the assumption
that all line stack entries are associated with atomic models, chiefly
because it had been tested with real spectral lines.

However, a number of entries appear in the stack that are not real lines
and trying to access their populations leads to a segfault.

Define method LinSv.isSpectralLine() to tell if a line is a genuine
spectral line, defined here as either a transferred (type 't') or
recombination (type 'r') line.  Use method to screen against other lines
(e.g., blends) when producing the command output.

Also use a check on whether the line transition is associated, to screen
against recombination lines, which are not associated with an atomic
model either.

Note that some recombination lines are transferred, so a test against
line type 'r' is not sufficient.

Finally, protect the population ratio in a safe_div() to guard against
zero populations (e.g., when the relevant ion has zero density).

Previously, the 'save line pop' command was described in a section, an
oversight that occurred  when renaming the old 'save hyperfine pop'
command to the current 'save line pop'.

Restored description to a subsection of the 'save line' section.

The previous description of the 'save line populations' command had
accidentally remained in Hazy 1.  Removed here.

The directory was accidentally added when the ca1_cmesh branch was
merged.  It was not intended for distribution with master.

Previously, there existed two commands for saving line populations:
'save line populations' reported the lower and upper level populations
for many (but not all) spectral lines known to the code, while
'save hyperfine populations' saved the populations and derived
quantities for a list of spectral lines (not necessarily hyperfine)
supplied by the user.

These are now brought together in a single scheme.  The command retains
its original name ('save line'), but the code it executes is that of the
new command ('save hyperfine').

The choice aims to simplify the output of the 'save line' command, which
produced large output files (several MB) even for single zone models.

Previously, when the command was used with the row keyword, it would
report, in one line of text, the ratio of level populations for all
spectral lines contained in the input file passed to the command.

Now the population ratio is still printed by default, but the individual
populations or the spin temperature may be reported, instead, by passing
an appropriate keyword.

This change is geared toward generalizing the command, and making more
useful, in preparation for merging it with the 'save line populations'
command.

Use GTH algorithm for all species with 0 source terms

See merge request !40

Set Fe8 default number of levels to 30 in all masterlists

See merge request !42

The Chianti10 Fe 8 model requires at least 30 levels so that the lines
at 131A and 270A be predicted.  The change was made on 2022-July-09 on
the CloudyChianti.ini masterlist, but the rest were omitted.

Here we update the remaining Cloudy* masterlists, retaining the original
date of change.

Update Chianti to v10

See merge request !39

Update Makefile for LLVM on Apple M1 chip

See merge request !38

The previous implementation contained several errors, which were
revealed when compiling on older versions of MacOS.

First, the CXX variable used was the one defined in source/Makefile, not
the one defined (later) in Makefile.conf.  Hoist the definition to the
top of the file to make sure we use the right compiler when setting the
system architecture.

Second, there was a space before "shell", i.e., "$( shell" was used
instead of "$(shell", which caused the command to **not** run in the
shell, and prevented acquisition of the LLVM target.  The variable was
never set, the string always evaluated to the empty string, and the M1
CPU was always picked up.  This caused warnings on older versions of
MacOS X (Yosemite (10.10), and Monterey (12.0)).  Remove extra space.

Third, there was an extra space at the end of the conditional, which is
appended to the end of the result of findstring.  On M1/M2 systems, this
is innocuous because the strings don't match anyway.  But on Intel
systems, findstring evaluates to an empty string, but the right-hand
side of the conditional to a single space, which does not match the
empty string on the left-hand side.  This causes the CPU to be set to
the unsupported apple-m1.  Compilation aborts with an error.  Rearrange
the conditional to use quoted strings instead of parentheses -- this
makes the test robust against spaces.  Also, rearrange the conditional
to use 'ifeq' instead of 'ifneq', and set the architecture to the
'native' default in an else clause.

FSF GCC complains that std::unary_function and std::binary_function are
deprecated as of the C++11 standard.  The latter has actually been
removed as of C++-17.

We'll deal with that later, when we upgrade to a more recent standard.
For now, silence the warning.

G++ v12.1.0 complains that a long cannot be safely printed in 2 digits;
the offending code updates grain labels.  Yet, the value printed always
fits in two digits, so no out-of-bounds memory access can occur at run
time.

Cast long value to unsigned short to silence the warning.

Compilation with G++ v12.1.0 on a MacOS X (v12.4) has brought up a few
new warnings that relate mostly to initialization.

Most of these occur in functions of thirdparty_quadpack.cpp.  Initialize
the offending floats therein to NaN, and long integers to 0.

Also, the long members of the CollSplinesArray class are now initialized
to 0.

Compilers for the M1 produce position-independent code by default.
Issuing '-fno-pie' on the command line emits an innocuous warning.

The present implementation checks on the compiler's capabilities only if
the chip is not arm64.

Clang++ version is:

   $  clang++ --version
   Apple clang version 13.0.0 (clang-1300.0.27.3)
   Target: arm64-apple-darwin21.5.0

Clang++ aborts when the '-march=native' compiler option is given.
Tailoring to the chip requires that '-mcpu=apple-m1' be given instead
in sys_llvm/Makefile.conf.

Clang++ version is:

  $ clang++ --version
  Apple clang version 13.0.0 (clang-1300.0.27.3)
  Target: arm64-apple-darwin21.5.0

The spin temperatures of these lines are not used on output, and can be
easily retrieved by calling HyperfineTspin() when needed.  Their storage
is unnecessary.

Chianti v10 contains transitions with the level indices swapped, meaning
that lower index > upper index.

Arrack checked only that the upper index be less than the maximum index
allowed for our AI cut.  In the case of fe_18, this let a few AI
transitions in the .wgfa file.

As a result, Cloudy crashed on a bounds-checking run.

Here, we remove fe_18.wgfa transitions with either index referring to AI
levels.  We also reverse the index order in some .wgfa and .splups files
so that the lower index appears first.

The line "Fe 8 270.090A" has been disabled in blr_n09_p22_Z20.in because
it does not exist in Chianti v10.

With the current setup, the test suite reports that it could not find
two Fe 8 lines at 130.941A and 270.090A used in agn_warm_absorber.in
and blr_n09_p22_Z20.in, respectively.

Inspection of the Chianti v7 fe_8.wgfa file reveals that these lines
correspond to transitions
  5 - 53   270.090   3s2.3p5.3d2 4D2.5 - 3s2.3p6.4f 2F2.5
  1 - 53   130.941   3s2.3p6.3d 2D1.5 - 3s2.3p6.4f 2F2.5

In Chianti v10, level 53 is indexed 64.

To recover these lines the default number of levels in CloudyChianti.ini
is here increased to 30 (the discrepancy being due to trimming by the
parser of levels with no experimental energy; the upper level index may
be found with the 'save lines labels' command).

Note that this recovers only the line at 130.941A.

The line at 270.090A is missing from the Chianti v10 data (fe_8.wgfa).

The file data/blends.ini and relevant infrastructure were introduced by
Peter van Hoof back in 2019 (2019-03-25 to 2019-04-09, according to the
SVN log).

However, the script fix_input_scripts_wl.pl was not updated to handle
updates to that file, partly because there had been no need for it.

The upgrade to Chianti v10 has required that the file be updated, and
the script is here modified to be able to carry such updates out.

It is also modified to handle PROBLEMs reported by Cloudy when, e.g., it
cannot identify a component to a line blend (a PROBLEM is followed by an
abort; updates must be done piecemeal).