Fix mass-consistent advection.

[p-costa]

This PR fixes and refactors the consistent advection implementation, which had issues in the calculation of the momentum and heat fluxes that led to stability problems. Other, unrelated minor updates have also been brought to this PR.

While unifying the conservative and non-conservative forms of the momentum equation, we have found some room for optimization and simplification of the implementation. Further steps could be taken that would lead to further simplification, but to keep things more efficient and avoid GPU register pressure, this is deemed to be enough for now.

Summary of changes:

• Updated cuDecomp;
• Updated minor change in the code block that computes the integrals of mass and energy;
• Changed dtmin -> dtmax in the input file;
• Skip operations when setting icheck/iout?d/isave to 0;
• Changed CONSERVATIVE_MOMENTUM to CONSISTENT_ADVECTION;
• Fixed mass--momentum-consistent advection discretization, and extended it for thermal energy (the latter is untested).

[gianlupo]

All looks good!