aggregator: jobs with collections on multiple backends

[jdries]

No description provided.

[soxofaan]

builds further on #95

[soxofaan]

already merged current state in master (feature has to be enabled with feature flag job_options={"split_strategy": "crossbackend"})

Still to do, follow-up features and related issues to look into:

• automatically detect need for and enable this cross-backend mode
• clarification: canonical (signed) batch job result URL + partial query param? openeo-api#507
• Batch jobs results: canonical link in canonical response? openeo-api#508
• load_stac with invalid/unreachable url: job stuck in status "created" openeo-geopyspark-driver#505
• merge_cubes: invalid spatial_extent merging openeo-geopyspark-driver#507
• current implementation only considers availability of collection ids, also handle process ids?
• smarter splitting than 'just split off each "alien" load_collection' node, make subgraphs largers?

[jdries]

reached initial goal, so closing!