Update some of our default values

[jdannberg]

This came out of a discussion at the ASPECT User Meeting. There are some default values that we have found out later could be chosen in a better way and that we might want to update. Let's collect all of those in this issue:

• [x] Change GMG to be the default solver #5211
• [x] For GMG: Increase number of Cheap solver steps/do not do expensive steps at all
• [ ] Particles: Increase Number of particles per cell per direction for Reference cell generator
• [ ] Particles: Is cell average the right default interpolator?
• [ ] Particles: Increase Minimum particles per cell?
• [ ] Should the default be to do no adaptive refinement? (Refinement + Coarsening fraction, Time steps between mesh refinement)

[tjhei]

• For GMG: Increase number of Cheap solver steps/do not do expensive steps at all

I would generally recommend choosing 0 expensive steps for GMG. That is different for the AMG solver, though.

With the current situation of the GMG not converging when switching from cheap to expensive, we could enforce either of the two values to be 0. @MFraters does that make sense to you?

[tjhei]

[ ] remove "Use conduction timestep" [ ] remove "Use direct solver for Stokes system"

[gassmoeller]

We did all the ASPECT 3 preparations that we could. This list is still useful to keep around and revisit for future releases, but everything that has not been done by now will have to wait for a future release.