feat(cli): Stop hardcode the reranker of choice. Implements #68

[Davidyz]

This PR refactored the reranker module to follow a similar paradigm as the embedding functions in chromadb. Users will now use the reranker key in the JSON config to specify the class (backend) for the reranker, and use the reranker_params field to initialise their chosen reranker class.

The original syntax, where users set the reranker model with the reranker key, is still supported, but will throw a warning message in the CLI and be removed in 0.6.0.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.93%. Comparing base (314cb01) to head (657bef4).
Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #85      +/-   ##
==========================================
+ Coverage   98.81%   98.93%   +0.11%     
==========================================
  Files          18       21       +3     
  Lines        1266     1312      +46     
==========================================
+ Hits         1251     1298      +47     
+ Misses         15       14       -1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Davidyz]

Should be pretty much done here. Still got some clean-up to do, and might try to simplify the rerank method so that it's easier to add support for new reranking backends.

[Davidyz]

Planning to add a method RerankerBase.compute_similarity(self, results: list[str], query_message: str) -> Sequence[float]. Concrete child classes would now implement this method that returns a list of similarity scores between the query message and each of the results, and RerankerBase.rerank, which is now a shared method for the base class, will assemble the final results. This saves tons of trouble when implementing a new reranker.

[Davidyz]

@ciekawy, any comments on my approach to the extensible rerankers? This base class will simplify the implementation of the reranker by taking care of building final scores for each document/chunk, and you'll only need to implement compute_similarity, which is a lot easier to work with.

EDIT: I should probably make this async, so that it'll work better with API-based reranker.