Remove compression argument from CF writer save_datasets

[BENR0]

Feature Request

Is your feature request related to a problem? Please describe.
The documentation says the compression argument is passed on to xarray's to_netcdf method and links to the xarray documentation for more information. This is a little confusing because
the to_netcdf method does not have a compression argument. Instead the compression can be specified b within the encoding dictionary. While this in itself might confuse users additionally it is unclear what will happen if the user supplies an encoding dictionary to the save_datasets method which includes information regarding compression. Which setting takes precedence then?

Describe the solution you'd like
In order to keep it close to the to_netcdf method the compression parameter should be removed and the documentation updated that compression is possible using the encoding parameter which will be passed to to_netcdf together with the link to the xarray documentation.

Describe any changes to existing user workflow
The default behavior should not change but users currently using the compression argument would need to change their code.

[sfinkens]

I agree, good catch!

[djhoese]

Do either of you have the time to make a PR for this?

[mraspaud]

@djhoese not really, but still...

[mraspaud]

Ok, after some discussion on slack, it seems to be a consensus that the default behaviour of compressing netcdf files should be phased-out to let the user take control through the encoding parameter and attribute, as is done in xarray.to_netcdf.