max_cloud_cover is ignored by Planetary Computer Sentinel‑2 data source

[robmarkcole]

Following examples in rslearn_projects I configured

            "data_source": {
                "cache_dir": "cache/planetary_computer",
                "ingest": false,
                "max_cloud_cover": 0,
                "name": "rslearn.data_sources.planetary_computer.Sentinel2",

But I still get images with clouds, and it appears max_cloud_cover is ignored. I think this should be addressed and errors raised on invalid args.

What is actually required is passing the cloud cover filter directly in the query:

            "data_source": {
                "cache_dir": "cache/planetary_computer",
                "ingest": false,
                "name": "rslearn.data_sources.planetary_computer.Sentinel2",
                "query": {"eo:cloud_cover": {"lte": 0}},

[favyen2]

This will be addressed by #371

In the example that was probably leftover from changing from using rslp.satlas.data_sources.MonthlyAzureSentinel2 (to get one image per calendar month, wrapping the Planetary Computer data source) to the new built-in PER_PERIOD_MOSAIC with rslearn.data_sources.planetary_computer.Sentinel2). The MonthlyAzureSentinel2 supported the max_cloud_cover option.