Refactor /nmdcschema/associations endpoint

[PeopleMakeCulture]

Spinoff from branch 401-sparql related to #401

[PeopleMakeCulture]

More consensus is needed on how this endpoint should work. There is a desire for something more generalized than a tool that navigates "has_input" and "has_output" relationships.

One idea is to use @eecavanna's refscan tool to create and cache a dict of possible paths between any two connected mongo collections.

I think this is something that will come up in the next referential integrity meeting.

In the meantime, I've prototyped a more specific use-case for traversing from data objects to biosamples [link to GH issue needed]

[shreddd]

implement with all-docs

[jeffbaumes]

@PeopleMakeCulture I think this related_ids() function is kind of how this endpoint could function. So a list of ids comes in, along with a target class type, and out comes an association between input ids and output ids. related_ids() outputs a list of pairs [input_id, output_id] which is a nice form for joining columns to data frames.

[dwinston]

some motivating context: folks are performing lots of client-side munging and writing logic that could be done server-side.