Add relationships functionality

[ml-evs]

Closes #71.

• Add subclassed relationships to include references key as in example spec 6.6.
• Populate included key in response with the relationships attached to each entry inside the collection.
• Actually make extra requests to populate included key in response with the reference's attributes, and validate them with reference models in the response.

A few notes/queries:

• I'm not entirely clear how we should approach populating the included key: should it be an additional request/request per reference in the server code, hardcoded to only look in the references endpoint? Similarly, to adhere to the JSON API spec, we'd need to provide endpoints such as /structures/mpf_1/relationships/references/author, which I can't see us implementing...
• I've had to change the JSON API implementation for relationship data from sets to lists as BaseResource is not hashable. As a result, the server code then has to loop over the relationships->references->data field to filter out duplicates based on ID.
• How should we handle validating that the relationship returned by the structure actually exists? JSON warnings?
• Do we want to support returning relationships as links rather than data (JSON API allows both/either,

[codecov]

Codecov Report

Merging #91 into master will increase coverage by 0.67%.
The diff coverage is 92.8%.

@@            Coverage Diff             @@
##           master      #91      +/-   ##
==========================================
+ Coverage   85.06%   85.74%   +0.67%     
==========================================
  Files          38       38              
  Lines        2210     2315     +105     
==========================================
+ Hits         1880     1985     +105     
  Misses        330      330

Impacted Files	Coverage Δ
optimade/models/links.py	86.95% <0%> (-3.96%)	⬇️
optimade/models/optimade_json.py	86.2% <0%> (ø)	⬆️
optimade/server/main.py	78.65% <100%> (+0.24%)	⬆️
optimade/models/tests/test_models.py	100% <100%> (ø)	⬆️
optimade/models/entries.py	100% <100%> (ø)	⬆️
optimade/server/utils.py	73.45% <100%> (+9.16%)	⬆️
optimade/server/tests/test_server.py	90.2% <100%> (+1.5%)	⬆️
optimade/server/mappers/structures.py	93.54% <100%> (+0.21%)	⬆️
optimade/validator/validator_model_patches.py	100% <100%> (ø)	⬆️
...ade/filtertransformers/tests/test_elasticsearch.py	80.64% <100%> (ø)	⬆️
... and 4 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 18181e2...b9290a5. Read the comment docs.

[CasperWA]

Looking good @ml-evs.

I think the only solution is to, indeed, do a second query for all collected relationships.
As I can see now, this is what is missing?

[CasperWA]

Haven't gone through your get_included_relationships function in detail, but it seems legit :)

[ml-evs]

I think this is basically ready now, the only thing that's not being done is returning a warning if the requested reference was not found under the references endpoint.

[CasperWA]

Seems to be in a state, where it can be merged in and we can call it "version 1" of adding relationships functionality.
Depending on how included and the include parameter will be defined finally in the spec, we will need to update this accordingly.

Thanks a bunch @ml-evs !

[ml-evs]

Seems to be in a state, where it can be merged in and we can call it "version 1" of adding relationships functionality.
Depending on how included and the include parameter will be defined finally in the spec, we will need to update this accordingly.

Thanks a bunch @ml-evs !

Yes, I came back to comment that this will currently populate included with any linked structures as well as references, which will probably need to be modified.

[CasperWA]

Thanks @ml-evs !