Fix errors in tests/MMS/elm-pb

[johnomotani]

Updates to be compatible with BOUT++ v5.0. Now compiles and runs.

Could do with more updates to remove deprecated functions, and possibly replace the Grad_parP method with the library version.

[johnomotani]

Test still running on Marconi to see if output is correct...

[johnomotani]

The test now runs, but does not pass. Errors are increasing if anything up to n=32

Bp0 = 4.444444e-02
psi width = 1.777778e-03
Making MMS elm-pb test
Grid file 'grid8.nc' already exists
  Running with  MZ=8 grid=grid8.nc
getmpirun: using the default mpirun -np
P : l-2 0.448528 l-inf 1.844185
Psi : l-2 0.007161 l-inf 0.018150
U : l-2 4.481064 l-inf 25.520997
Grid file 'grid16.nc' already exists
  Running with  MZ=16 grid=grid16.nc
getmpirun: using the default mpirun -np
P : l-2 0.361313 l-inf 2.253608
Psi : l-2 0.007330 l-inf 0.019226
U : l-2 8.755153 l-inf 52.830942
Grid file 'grid32.nc' already exists
  Running with  MZ=32 grid=grid32.nc
getmpirun: using the default mpirun -np
P : l-2 0.406318 l-inf 2.436382
Psi : l-2 0.007367 l-inf 0.019666
U : l-2 11.800941 l-inf 116.680718
Grid file 'grid64.nc' already exists
  Running with  MZ=64 grid=grid64.nc
getmpirun: using the default mpirun -np

and the n=64 case fails to complete with At t = 0.000530184, mxstep steps taken before reaching tout.
Could fix that error, but already takes 5 hours to run the test as far as I did, and I increased the processor numbers used up to 128. So I think this test needs more attention than I have time to give it...

[ZedThree]

These tests are both tricky and fragile. I think these changes are worth merging at any rate