These protect against division by zero.  Problems were exposed
by the lab sim posted in gtests / bugs / Mancini.  The
sim does get now much farther along but aborts when the line
stack becomes misaligned.

The sim is a lab plasma, so it is far from our home base, but we
should be able to compute it.  It is non-equilibrium.  It is
unclear why the cooling is zero - perhaps the total is net heating.

The changes seem safe since they are have never been needed and
will have no effect in most cases.