Add test for anemoi-datasets loader using CERRA test dataset hosted on EWC

[leifdenby]

I have added a test that applies the anemoi-datasets loader and then uses mlwp-data-specs validator to validate the loaded xr.Dataset.

To implement this I also uploaded cerra-rr-an-oper-0001-mars-5p5km-2017-2017-6h-v3-testing.zarr from the anemoi dataset registry to the EWC (European Weather Cloud) object store at https://object-store.os-api.cci2.ecmwf.int/mlwp-sample-datasets/anemoi-datasets/cerra-rr-an-oper-0001-mars-5p5km-2017-2017-6h-v3-testing.zarr/

The point of this PR is to adjust mlwp-data-specs and the anemoi-datasets loader here in mlwp-data-loaders until the validation of the loaded xr.Dataset passes.

Currently this installs mlwp-data-specs directly from main (i.e. whatever commit is on HEAD) of https://github.com/mlwp-tools/mlwp-data-specs. Eventually we should tag and publish a release of mlwp-data-specs once this data-loader (and one for anemoi-inference datasets?`) pass the validation.

@mpvginde your thoughts would be much appreciated :)

[leifdenby]

Currently the test fails, I need to find out if that is because either:

1. the loader isn't implemented correctly, i.e. I missed something when refactoring the loader functionality from mxalign
2. the validator isn't implemented correctly, again I could have missed something from mxalign

[mpvginde]

Hi @leifdenby from my own tests with using the validator with the mxalign loaders I have noticed that some of the attributes (long_name, units, ..) are not currently not set correctly or not provided at all by the loaders. This is probably causing the validation to fail

[leifdenby]

I've (temporarily) integrated mxaligns validation and the validation passes (in the tests) using both mxalign and the validation logic currently (as of https://github.com/mlwp-tools/mlwp-data-specs/tree/b703bacd89854441316fb36c321aa03757ee233e, current HEAD of main) in mlwp-data-specs 🥳

Given that the loaded dataset now passes both from the perspective of mxalign and mlwp-data-specs I am going to merge this PR :)