Improve use of CCPP I/O reader in gravity wave drag schemes

From review of GW https://github.com/ESCOMP/atmospheric_physics/pull/292

For moving mountain
```fortran
! Allocate mfcc. "desc%maxh" and "desc%maxuh" are from the file, but the
! model determines wavenumber dimension.
allocate (desc%mfcc(desc%maxh, -desc%maxuh:desc%maxuh, &
-band%ngwv:band%ngwv), stat=errflg, errmsg=errmsg)
if(errflg /= 0) then
return
endif

! Get mfcc data.
call reader%get_var('NEWMF', file_mfcc, errmsg, errflg)
if (errflg /= 0) then
return
end if

! original formulation is
! desc%mfcc(:,-desc%maxuh:desc%maxuh,-band%ngwv:band%ngwv) = file_mfcc(:,:,ngwv_file-band%ngwv+1:)
! where band%ngwv = 0 -- this cannot be handled directly by the
! ccpp i/o reader, so copy it in from the temporary array.
desc%mfcc(:, -desc%maxuh:desc%maxuh, 0) = file_mfcc(:, :)
```
@nusbaume wrote:
> It might be good to eventually figure out what exactly can't be handled by the pio_reader here (maybe worth including in the "future work" issue?).

For Beres/convective:
```fortran
call reader%get_var('MW', tmp_var1d, errmsg, errflg)
if (errflg /= 0) then
return
end if
desc%maxuh = size(tmp_var1d)
deallocate(tmp_var1d, stat=errflg)
```
@nusbaume wrote:
> It would be good to replace this and the code below with a get_dim method at some point. Maybe worth adding to the "future work" issue?

@peverwhee wrote:
> can we get these from the file instead of hard-coding them? Maybe this is part of https://github.com/ESCOMP/atmospheric_physics/issues/321 ?

for the moving mountain initialization code as well.