Issue when using different CRS in load_collection and aggregate_spatial

Issue reported by colleague Landuyt, using a geometry initially defined in UTM, but ending up in different CRSes in process graph.

e.g. EPSG:32631 in load_collection spatial extent and implicit EPSG:4326 in aggregate_spatial geometry:

{
  "process_graph": {
    "loadcollection1": {
      "process_id": "load_collection",
      "arguments": {
        "bands": ["B02"],
        "id": "TERRASCOPE_S2_TOC_V2",
        "spatial_extent": {"west": 476580, "south": 5641538, "east": 492312, "north": 5650941, "crs": "EPSG:32631"},
        "temporal_extent": ["2022-09-01", "2022-09-15"]
      }
    },
    "aggregatespatial1": {
      "process_id": "aggregate_spatial",
      "arguments": {
        "data": {"from_node": "loadcollection1"},
        "geometries": {
          "type": "FeatureCollection",
          "features": [
            {
              "type": "Feature",
              "properties": {},
              "geometry": {
                "type": "Polygon",
                "coordinates": [
                  [
                    [2.8906121198012946, 50.9254284428713],
                    [2.890413345052933, 51.00998523843394],
                    [2.6661682282259633, 51.0095597508646],
                    [2.6667737398200884, 50.925004228736995],
                    [2.8906121198012946, 50.9254284428713]
                  ]
                ]
              }
            }
          ]
        },
        "reducer": {
          "process_graph": {
            "mean1": {"process_id": "mean", "arguments": {"data": {"from_parameter": "data"}}, "result": true}
          }
        }
      },
      "result": true
    }
  }
}

Fails currently with

OpenEoApiError: [500] Internal: Server error: RuntimeError("Unexpected crs: '+proj=utm +zone=31 +datum=wgs84 +units=m +no_defs +type=crs' != '+proj=longlat +datum=wgs84 +no_defs +type=crs'") (ref: r-d61de60afce04da99952ec268167beec)

bottom part of related stack trace:

  File "/opt/venv/lib64/python3.8/site-packages/openeo_driver/ProcessGraphDeserializer.py", line 1484, in apply_process
    return process_function(args=ProcessArgs(args, process_id=process_id), env=env)
  File "/opt/venv/lib64/python3.8/site-packages/openeo_driver/ProcessGraphDeserializer.py", line 552, in load_collection
    return env.backend_implementation.catalog.load_collection(collection_id, load_params=load_params, env=env)
  File "/opt/venv/lib64/python3.8/site-packages/openeo/util.py", line 386, in wrapper
    return f(*args, **kwargs)
  File "/opt/venv/lib64/python3.8/site-packages/openeogeotrellis/layercatalog.py", line 123, in load_collection
    return self._load_collection_cached(collection_id, load_params, WhiteListEvalEnv(env,WHITELIST))
  File "/opt/venv/lib64/python3.8/site-packages/openeogeotrellis/layercatalog.py", line 230, in _load_collection_cached
    projected_polygons = to_projected_polygons(
  File "/opt/venv/lib64/python3.8/site-packages/openeogeotrellis/utils.py", line 203, in to_projected_polygons
    raise RuntimeError(f"Unexpected crs: {provided_crs!r} != {expected_crs!r}")
RuntimeError: Unexpected crs: '+proj=utm +zone=31 +datum=wgs84 +units=m +no_defs +type=crs' != '+proj=longlat +datum=wgs84 +no_defs +type=crs'

Reproduce the supplied process graph, then start in openeogeotrellis/utils.py at to_projected_polygons and trace the call from openeogeotrellis/layercatalog.py in load_collection. Use the reported CRS mismatch and stack trace to understand the failure path; done means the example no longer fails because load_collection and aggregate_spatial use different CRSes.