run_udf: add tests for case inside callback

[jdries]

These tests show how a user would typically use run_udf.

They also seem to expose a remaining issue in current implementation, in the sense that the UDF continues to receive the full input datacube as opposed to only receiving a chunk.

Forcing parallellization with dask does reveal error messages, so this is probably what we need to get behaviour similar to a 'real' backend.

[ValentinaHutter]

Pre-commit should be included in your poetry setup, but you can also resolve the pre-commit hook issue manually by running: pre-commit run --all-files

https://pre-commit.com/