basins_stats.compute fails with basins_gdf argument

[jsta]

Even with a tiny polygon object, the earthengine API says:

ee.ee_exception.EEException: Object too large (4000000016 bytes).

You can pass an asset path with the gee_feature_path argument to get around this error.

[jonschwenk]

This could have happened because I was maxxing out rabpro's GEE memory with other runs. Those have been cancelled so maybe try again? Or if you were using your personal GEE account then it's a real bug.

[jsta]

I was using my personal account.

[jsta]

Works:

• all basin stats queries using a vector asset already on GEE

• all basin stats queries using a time_stats reducer

Doesn't work:

• uploading a feature asset on-the-fly but only with specific image raster assets

# works
urls, tasks = rabpro.basin_stats.compute(
        [Dataset("projects/soilgrids-isric/soc_mean", "soc_0-5cm_mean", stats=statlist, gee_type="image")], 
    basins_gdf=gdf, validate_dataset_list=False)

# doesn't work, doesn't matter what the resolution is set at
urls, tasks = rabpro.basin_stats.compute(
        [Dataset("JRC/GSW1_3/GlobalSurfaceWater", "occurrence", stats=statlist, gee_type="image")], 
    basins_gdf=gdf, validate_dataset_list=False)

Proposed fix:

Arbitrarily subject image assets to a time reducer which should have no effect other than avoiding the above error