Consider Dirichlet / sponge for dycore upper boundaries

* Current PAM-A allows free changes of the upper atmosphere in the model (no knowledge of GCM data past CRM extent)
* GCMs expect an upper atmosphere similar to the GCM column. Violations of this lead to bad CRM forcing to the GCM.
* Consider setting upper boundary ghost cells to GCM column values
* Will also need a sponge to avoid reflection of internal waves back into the model
* This sponge will be a **nudging** sponge like you see in regional model lateral boundaries, not like SAM's / PAM's existing sponge
* Concern: Will this cause problems? Energy conservation? Other issues?
* This **should** fix Walter's issue he's currently seeing in PAM (A-grid) + SCAM

@whannah1 , @celdredsandia , @mwarusz , @xyuan , @mt5555 , please give your thoughts if you have any regarding this.