hardcode ZENODO_FILES_DATA in WorldCerealConfidences

[gabrieltseng]

This prevents the following:

requests.exceptions.HTTPError: 429 Client Error: TOO MANY REQUESTS for url: https://zenodo.org/api/deposit/depositions/7875105/files

Since any changes to the files changes the ZENODO_RECORD_ID, I think this is safe to do.

[favyen2]

In other data sources that require something similar, we cache the data to a file and then read it in the parallel workers so that as long as --no-use-initial-job is NOT passed then it will save the file on the initial job and then afterward they workers will just read from the file.

For example, the first call to LocalFiles.list_items writes a summary.json showing the bounds of all of the raster files and then successive calls can read that instead of opening a lot of GeoTIFFs to get the metadata.

I think it is a bit different here since there are already a fixed set of files you are expecting, so I think it is fine to hardcode if you prefer, but I would cut it down to just the minimal amount of data we need (filename and download link) instead of keeping the entire response.

[favyen2]

looks good to me!