Framework for writing parameters to netCDF format

[jkrasting]

This PR introduces a framework to write useful parameters to a netCDF file. Several of us have run into instances where offline analyses and scripts have used hard-coded values for Rho0 and Cp that differ from those used in the model simulation. The introduction of this “ocean_parameters.nc” file will help ameliorate some of those issues and allow scripts/analyses to load these values on a per-experiment basis.

@StephenGriffies also specifies CMOR names and attributes for these in the 2016 OMIP description paper - specifically for ‘rhozero’ and ‘cpocean’. These attributes are included in the register_diag calls.

In this commit:

• Added write_parameter_fields subroutine to MOM.F90
• Makes use of the FMS Diag Manager "scalar_axis", which
  is added as axes%axesNull

And a sample diag_table used for testing:

"ocean_model", "C_p", "C_p", "ocean_parameters", "all", .false., "none", 1 
"ocean_model", "Rho_0", "Rho_0"," ocean_parameters", "all", .false., "none", 1 
"ocean_model", "cpocean", "cpocean", "ocean_parameters", "all", .false., "none", 1 
"ocean_model", "rhozero", "rhozero", "ocean_parameters", "all", .false., "none", 1

[Hallberg-NOAA]

Thank you, @jkrasting, for this useful addition.

Over time, we collectively seem to have acquired the unfortunate habit of adding a great many diagnostics directly to MOM.F90, which has now become severely bloated at more than 4200 lines with almost a hundred diagostics, obfuscating the intent of the code and severely inhibiting any future top-level restructuring of the MOM6 code. This particular PR is a good example of something that need not be in MOM.F90, and could easily be added to one of the files in the diagnostics sub-directory.

In the future, any diagnostics added directly to MOM.F90 should only occur with the greatest deliberation and a really good justification why they could not be added somewhere lower in the MOM6 code tree.

[jkrasting]

Yes, we've discussed this before @Hallberg-NOAA.

Despite your legitimate concerns about excess diagnostics in MOM.F90, I included this new code here for two reasons:

1. The new ocean_parameters.nc file is extremely similar in concept to the ocean_static.nc file. The code for both files should be co-located.
2. The parameters should be limited to broad, cross-cutting parameters - such as Cp, Rho, gravity, etc. We should be relatively juducious about which parameters to include in this file. It should NOT be a duplicate of the MOM_parameter.doc file, but it does need to include parameters that are relevant to offline analysis and are prone to changing with different experiments.