Better error running basin_stats with a data_id not in catalog metadata

[jsta]

Consider this scenario:

• a data source like GlobalSurfaceWater Dataset comes out with a new version
• the catalog watcher removes the old version from the catalog (it still exists on GEE)
• if you run a query against the old version, you will get a cryptic error (preceded way up the stack by a warning that the id is not in the catalog)

This is because the call to Dataset emits a warning (not an error) and passes on an empty output to basin_stats.compute because the id is not in the catalog (I think this is fine). compute attempts to proceed as normal with the "empty" request (I think this is not fine). We should probably catch these in compute and throw an error.

[jonschwenk]

Should be easy enough to implement.

For this particular case (and maybe others like it), I think the solution would be to put deprecated assets in rabpro's user_added_catalog. This would avoid errors but this is probably a case where you want it to fail (if you weren't aware of the new version) so maybe that's a bad idea.

[jsta]

Should we enable users to override any checks against the catalog and just run the query? For reasons that are unclear to me, DAYMET does not appear in the scraped json but is clearly available in the GEE catalog.

[jonschwenk]

Yes, I think so. I was struggling with this for VotE forcings/characteristics fetching as well.

[jsta]

The two dataset attributes that you get "for free" if your entry is in the catalog is type and resolution. I've made it so you can skip dataset validation (with validate_dataset_list=False) and still execute your query by manually specifying these values.