Convert NWP grib files to Zarr intermediate

[JackKelly]

Detailed Description

#122 found that it'll be way too slow to load NWPs from native files directly. So we need to convert to Zarr first.

All of "file 3" listed in this Met Office PDF look useful for helping to predict the evolution of clouds.

Related issues:

• Check for -1 values in sat data #29
• Re-create NWP Zarr with one chunk per init_time and step and small bit depth #30
• Re-chunk Satellite Zarr with one chunk for all channels and full spatial extent #44
• Experiment with ingesting NWP data from raw files, including relevant params at cloud heights #122
• Download additional UK Met Office UKV NWPs #196
• Write script to convert .nat EUMETSAT files to Zarr intermediate Satip#14

Also see:

• Met Office's data docs
• https://github.com/JackKelly/ng_nwp/blob/master/extract_grib.ipynb
• OCF's data docs on NWP data
• rechunk_nwp_data.py

Something like this should work:

ds = xr.open_dataset(
    filename,
    engine="cfgrib",

    # Wind direction and wind speed are the only variables which are at 
    # 10 meters above ground in the UKV GRIB files.
    #backend_kwargs={'filter_by_keys': {'level': 10}}
    backend_kwargs=dict(filter_by_keys={'stepType': 'instant'})
)

And load twice: Once with stepType=instant and one with stepType=accum. And watch for all the fun exceptions that cfgrib throws :)

This looks like it'll be quite slow. So definitely need to parallelise. If it's super-slow then might be best to FTP new grib files to the cloud and do the processing in parallel in the cloud.