Dynamic h-refinement in BBH evolution

[nilsvu]

We want to periodically evaluate h-refinement criteria throughout the evolution. This can be limited to slab boundaries but needs to work with LTS. This can be treated separately from p-refinement, which should run more frequently.

1. At some slab boundary switch to AMR phase
2. Evaluate h-refinement criteria, split and join elements, go back to evolution phase

The main thing that's missing for this to work is the projection of the time stepper history during h-refinement (GTS and LTS).