GRIB 2 Output

[clechartre]

Purpose

Outputting the inference results as GRIB format. This integrates the functionality of transforming the .npy array that is originally produced into a GRIB from a template.

Code changes:

• ar_model: add to the function on_predict_epoch_end a function generate_time_step to generate time steps markers for the grib files and a function that translates the npy output into a grib file.
• other modifications to the code include adaptation to tsa, imports of new libraries for this functionality, and references to templates in constants

Checklist

Before submitting this PR, please make sure:

• [x] You have followed the coding standards guidelines established at Code Review Checklist.
• [x] Docstrings and type hints are added to new and updated routines, as appropriate
• [x] All relevant documentation has been updated or added (e.g. README)

Review

For the review process follow the guidelines at Checklist

[joeloskarsson]

I think this might require a bit more discussion before implementing in main. I agree that the numpy format that example forecasts are output to is not optimal. My view would be instead go to xarray+zarr, to keep to the same format as the input data. Then you could convert that to GRIB if you would like, but that would be the job of another tool outside the scope of this repo.