Design patch-to-point job definition

[kvantricht]

Once our extractions are migrated to STAC api, we need to think about how to launch actual patch-to-point jobs, most likely per ref_id, then further split into per EPSG zone.

[VincentVerelst]

A first draft, very prone to change:

• Query STAC API with a property filter equal to a certain EPSG code
• From this we can construct a new job_tracker where each job corresponds to a different EPSG code
• Get the total bounds of all STAC items returned from the query. With these bounds query the RDM API to get all features intersecting with the bounds and with extract_flag==1 --> Here we need a sampling procedure (just take centroids?)
• load_stac the S1 and S2 patch extraction from the API and load pre-composited Agera5 and DEM, apply standard preprocessing and merge
• aggregate_spatial on the sampled points and execute the batch job --> To be seen if it's possible in one batch job
• Download results as geoparquet and merge to existing geoparquet database of point extractions
• If job succesful: mark all extracted features as extracted (using extract_flag?) in RDM, to avoid double extractions