resample_(cube_)spatial: support metadata transform

[jdries]

The cube metadata classes currently do not support resampling the spatial dimension. As a consequence, this metadata is incorrect and misaligned with the actual state on the server.
Server-side, the same problem exists in our 'dry-run' setup: no resampling, so incorrect metadata.

By implementing this in this repo, we can solve both problems at once.
For the user, better client metadata for cubes gives another means to detect issues early.

[VictorVerhaert]

This issue (and mainly its related fix to the backend) is relevant for EUGW as one pipeline produces incorrect results when reprojecting

[soxofaan]

Is there any pointer to a concrete use case or snippet where this goes wrong?

[VictorVerhaert]

Here is a job-id that gives incorrect results: j-2501091523364ccc8d462ddee1e5b385

[soxofaan]

can you also comment on what is wrong with the result (what you get versus what you expect)?

[JeroenVerstraelen]

Open-EO/openeo-python-driver#348

[VictorVerhaert]

What I get: (spatial extent in latlon followed by resample to 3035)

What I expect: (spatial extent in 3035 followed by resample to 3035)

(view is in EPSG:3035)

[VictorVerhaert]

python code to recreate the above images:

import openeo
import pyproj
from pathlib import Path

# establish connection to OpenEO backend
connection = (openeo.connect("openeo.dataspace.copernicus.eu"))
connection.authenticate_oidc()

root_dir = Path(__file__).parent / "base_case"
if not root_dir.exists():
    root_dir.mkdir()

# define start date and end date
start_date = "2022-01-01"
end_date = "2022-12-31"

def reproject_extent(extent, from_epsg, to_epsg):
    transformer = pyproj.Transformer.from_crs(f"epsg:{from_epsg}", f"epsg:{to_epsg}", always_xy=True)
    west, south = transformer.transform(extent['west'], extent['south'])
    east, north = transformer.transform(extent['east'], extent['north'])
    return {'west': west, 'south': south, 'east': east, 'north': north, 'crs': f"epsg:{to_epsg}"}

original_spatial_extent = {'west': 19.893982316864847, 'south': 49.16379666241664, 'east': 20.125833976862033, 'north': 49.31558120075986}
reprojected_spatial_extent = reproject_extent(original_spatial_extent, 4326, 3035)

cube_original_crs = connection.load_collection(
        collection_id="COPERNICUS_VEGETATION_PHENOLOGY_PRODUCTIVITY_10M_SEASON1",
        temporal_extent=[start_date, end_date],
        bands=["SPROD"],
        spatial_extent=original_spatial_extent
).resample_spatial(resolution=10, projection=3035, method="bilinear")
cube_original_crs.download(root_dir / "cube_original_crs.tif")

cube_projected_crs = connection.load_collection(
        collection_id="COPERNICUS_VEGETATION_PHENOLOGY_PRODUCTIVITY_10M_SEASON1",
        temporal_extent=[start_date, end_date],
        bands=["SPROD"],
        spatial_extent=reprojected_spatial_extent
).resample_spatial(resolution=10, projection=3035, method="bilinear")
cube_projected_crs.download(root_dir / "cube_projected_crs.tif")

[soxofaan]

started with pull request #712