Decide whether the AMR callback should use the number of total or accepted steps

[ranocha]

See #771, #759, #769, #19

[ranocha]

@sloede?

[sloede]

I'd say to use the number of accepted steps. Why?

1. Principle of least astonishment (see also Weird screen output behaviour: alive and analyse "forgotten"?  #759)
2. AMR is essentially a performance optimization: Given the solution changes over time, adapt the mesh accordingly. Since each discarded time step does not change the solution, it should not be counted towards the "time" between two adaptations.

But maybe I forget something, so I'd like to get the opinion of @gregorgassner or @amrueda on this, who use AMR more frequently.

[ranocha]

I'd say to use the number of accepted steps

I think that, too. It just requires a bit of work and manual checking since the results of at least one elixir will change slightly

[amrueda]

I agree, AMR should use the number of accepted steps, as the solution does not change in the discarded steps.