Some introspection about molecular structure

In his review of #439, @pdmullen pointed out that it would be nice to initialize an ideal gas with, e.g.,

IdealGas(Real gm1, MeanAtomicProperties prop)

and compute Cv automatically from that. But this would require knowledge of, e.g., the mean molecular weight $\mu$, rather than the mean atomic mass.

This may also be relevant for more detailed 3T models. We should consider a strategy to enable this down the line. Perhaps, for example, by storing the number of atoms in a molecule in the MeanAtomicProperties struct, or by passing this in to the ideal gas constructor.