Adding "physics-based" reconstruction

[BenWibking]

For various things, it would be useful to reconstruct a different set of variables other than the primitive variables. For example, we might want to reconstruct (density, velocity, temperature) instead, or we might we to reconstruct using the characteristic waves.

How can this best be accomplished with the current machinery? Can we do this cell-by-cell, e.g., with an inline function within the reconstruction kernel?

[pgrete]

Short answer: yes, that should be possible

Depending on many different variations this will get, it'd make either sense to extend the current calculate flux function to do some conversion before calling the reconstruction functions or duplicate the calculate flux kernel if a lot more logic would be required).
The conversion should be straightforward, once the individual cell based inline conversion functions that @forrestglines implemented in the cluster branch are in place.

[BenWibking]

Reconstructing in temperature instead of pressure is necessary for my current simulation project. Will follow up to see if we can discuss this soon.