implement a small_density cutoff

[zingale]

For the compressible problems, when modeling, e.g. atmospheres, we can run into a situation where the density might go negative. We should have a density floor as a runtime parameter.

This should probably be done after each stage update for the RK/SDC solvers.