Export Netcdf file

[gmaze]

As of now, Argopy has focused on accessing/reading/loading Argo data as xarray datasets.

But for operators who modify the content of a dataset, by adding calibration information for instance, there is a need to be able export these updated/augmented xarray datasets to netcdf files, following the Argo conventions for variables encoding.

This should requires a static dictionary asset with all possible Argo variables and the corresponding encoding information.

One tricky part is for string variables.
For instance:DATA_TYPE is encoded as an array of dtype('S1') with length 16, along a STRING16 dimensions, in netcdf files
Hence, when decoded abruptly with xarray, it is returned as 16 values with dtype=|S1], along a STRING16 dimensions of length 16.
But after variable casting by argopy, this variable is more appropriately returned as 1 value with dtype=<U16

A possible API could be:

# Load raw data (but appropriately casted by argopy):
ds = ArgoFloat(WMO).open_dataset('prof')

# [...] Do some stuff on ds 

# Export dataset as a clean Argo netcdf file:
ds.argo.to_netcdf('updated.nc', convention='Argo-3.1 CF-1.6')

The goal would be for the new netcdf to pass the Argo file checker and be valid for submission to a DAC

[gmaze]

poke @CKermabon @quai20 @tcarval

[gmaze]

there is a lot of features in the bgcArgoDMQC library in here:
https://github.com/ArgoCanada/bgcArgoDMQC/blob/main/bgcArgoDMQC/io/netcdf.py

but it is all based on low-level netcdf4 dataset manipulation