GDMP sync result performance

[jdries]

13:42:03.496 Received request 13:42:04.269 Loading with params DataCubeParameters( 13:42:10.726 Cube partitioner index: org.openeo.geotrelliscommon.package$ConfigurableSpaceTimePartitioner@3f0ed609 13:42:11.141 SaveRDD 13:42:23.020 +20s Writing geotiff to /tmp/openeo-pydrvr-bl4mwpa5.save_result.gtiff with type float32 and bands 1.0

13:42:35.095 +32s Request fully done

Two observations:

• load_collection consumes 6 seconds at some point, perhaps for looking up the netcdf files on disk?
• there's too many partitions: we know input dates, so it should be feasible to construct a sparse partitioner very fast
• reduction over time could be optimized, but most of the time is spent loading the data

[jdries]

partitions are now lower on openeo-dev, job runs with 1 stage as well most of the time is still spent loading data

[jdries]

New GDMP_TEST layer still has an error because we try to interpret netcdf's as geotiff:

▸Server error: geotrellis.raster.io.geotiff.reader.MalformedGeoTiffException: incorrect byte order
    error
    Code: Internal
    ID: r-2401162411bd4614be228d2aace68462

[jdries]

committed collection test for cgls, avoid need for a custom OpensearchClient

[jdries]

test collection now works, there was a pixel shift, so I set realign to true

[jdries]

it works now, next step: remove test layer and move config into 'real' layer

[jdries]

committed collection changes, to be communicated to users