Anelastic cutoff issues in Urca sims

[biboyd]

There appears to be some sort of perturbation that arises at the anelastic_cutoff_density. This grows and starts to limit the timestepping if the interior convection is relatively slow. Example slices of the density perturbation and resulting velocity:

The provided inputs and model file can replicate this for a 256^3 inputs_model.zip

I've tested this using pure MPI and MPI+OMP, with the same results.