Add some caching to AggregatorSecondaryServices.service_types

[JohanKJSchreurs]

Add some caching to AggregatorSecondaryServices.service_types.

This is a follow up ticket for the to do's of #78, to plan and track the tasks separately .
See also: #78 (comment)

[soxofaan]

What also can be covered in this ticket is another todo from #78 (comment):

determining which upstream backend to pick should (also) be based on service type

[soxofaan]

some more background/explanation:

• you can work with assumption "there is only one upstream backend per service type" (e.g. XYZ -> SentinelHub at the moment) so deciding which upstream backend to use can be done purely based on service type. There is also no need to implement "merging" of secondary service information. This assumption will not hold in the long term (depends on discussion Federation of Secondary Services #83) but is good enough for now. Just put "TODO" comment wherever that makes sense
• don't just purely cache the return value of AggregatorSecondaryServices.service_types. Instead build (and cache) an internal housekeeping structure and use that to:
  • extract AggregatorSecondaryServices.service_types response
  • use it in AggregatorSecondaryServices.create_service to determine upstream backend from given service_type. So replace

    openeo-aggregator/src/openeo_aggregator/backend.py

    Lines 751 to 760 in effd2a9

    	# TODO: hardcoded/forced "SentinelHub only" support for now.
    	# Instead, properly determine backend based on service type?
    	# See https://github.com/Open-EO/openeo-aggregator/issues/78#issuecomment-1326180557
    	# and https://github.com/Open-EO/openeo-aggregator/issues/83
    	if "sentinelhub" in self._backends._backend_urls:
    	backend_id = "sentinelhub"
    	else:
    	backend_id = self._processing.get_backend_for_process_graph(
    	process_graph=process_graph, api_version=api_version
    	)

• for the internal housekeeping structure: just use nested dicts/lists, so that serialization in caching layer (e.g. byte strings for zookeeper backend) works without a lot of additional work.
• for the caching: look for "_memoizer" and "memoizer_from_config" usage in backend.py for inspiration how caching is done in other subsystems there. For example: AggregatorCollectionCatalog._get_all_metadata_cached is also a caching layer around a single internal housekeeping structure that is used to extract different derivative products

[soxofaan]

merged #86