consider removing reference states from eigenvectors

[zingale]

Currently, in the ppm tracing, we use the reference states in computing the eigenvectors. But this means that, since we include sources now in the reference states, that all the fluid variable states are essentially linked to the source terms. And if the source terms are not well-behaved, then all states are affected.

This is one of the reasons why the problem that prompts #1962 can arise: with SDC, the pressure source term can be large and negative, resulting in the pressure going negative and the reference pressure being reset to small_pres, and then this influencing density through the various alphas.

[zingale]

Note: using the reference states in the eigenvectors was never explicitly addressed in the PPM paper. It just says "modified slightly for the present application" just before Eq. 3.6, which is the first place the tildes appear in the eigenvector products.