ISMR

[bprather]

This is an attempt to split the internal mesh refinement pieces out of feature/multizone_merge. It is still experimental -- in particular, it does not yet recalculate the floors and sound speeds for zones in the de-refinement region, and thus has some trouble integrating difficult states without high floors and other special accommodations. Since the scheme is first-order, it also has a stability factor of 0.5 rather than the usual 1.0 for second-order schemes. However, with a few layers of ISMR the time-step of spherical simulations can be dramatically improved, at the cost of some accuracy right next to the coordinate pole.

Please note this is mostly the work of @hyerincho! I'm just backporting it out from the other Multi-zone changes she's made, since it might prove broadly useful. In light of that, if you do make use of this code, please credit her, either via looping her into ensuing projects/papers, or by citing the implementation description to be published this fall.