Bug: OBS does not seem to take into account the eos used in NEMO

[mathiot]

Context

OBS interpolates NEMO T and S onto the obs location. It works from the EN4 data converted into feedback files. These observations are in potential temperature and practical salinity. After looking at the model data interpolated and the code I think the data are not converted to the observations eos and the netcdf attributes of the _Hx variables are missleading.

Analysis

I see different possible answer to improve this:

• convert model T and S to observation expected T and S based on model eos and observation eos (function for conversion available in eosbn2 I think)
• do not convert the model T and S but adapt the netcdf attributes of _Hx variables (not convinced it is in the spirit of the OBS operator)

Fix

For solution 1, a call to eos_pt_from_ct can do the job for temperature. For salinity, a function eos_ps_from_sa need to be implemented.