Extend and improve process graph repository

[christophfriedrich]

There is a process graph repository, but it is very basic. It could be improved by the following things:

• The Hub could also list all back-ends that support the processes in the process graph (idea by @m-mohr taken from Add process graph repository #2)
• Add search feature for submitted process graphs -- once the repository contains more than a handful of entries, keeping them discoverable is important ✳️
• Add user management to process graph exchange -- currently, the repository only allows the creation, but not the modification or deletion of submitted entries; and, maybe even more importantly, snippets are not attributed to their authors ✳️
• Add endpoint that returns individual submitted process graphs by identifier -- this would allow users to seamlessly use snippets from the openEO Hub repository in their own process graphs via openEO's run_process_graph process ✳️
• Make the process graph repository a 'pseudo-backend' by implementing the openEO API's /process_graphs endpoints -- this way, users could interact with the openEO Hub repository via the openEO-API-compatible tools they already know and use ✳️

list to be continued

✳️ = idea taken from my bachelor thesis

See #2 for the initial process graph repository idea.

[christophfriedrich]

Point 4 in the list above is essentially a sub-feature of point 5: GET /process_graphs/{process_graph_id} is one of the five /process_graphs endpoints.

Commit a6c2e5e made GET /process_graphs and POST /process_graphs openEO-API-compliant (v0.4.2); and in dd3d267 I added GET /process_graphs/{process_graph_id}.

The PATCH and DELETE endpoints shouldn't be supported before there is no user management, otherwise anyone could just remove all stored process graphs. I.e. point 5 depends on point 3.

➡️ Point 4 is done, point 5 is now supported 3/5 = 60%.

[m-mohr]

This issue is on hold and needs further investigation based on the outcomes of Open-EO/openeo-api#85 for process graphs.

[bgoesswe]

Just to add:
I tried to use this feature for the QGIS plugin, but for better user experience, it would be nice to know on which backend the process graph was originally executed, just to know where it should work unchanged and where to get the info about the collection id used.