Join node should allow reciprocal rank fusion as additional merging method

[mathislucka]

Proposed changes:

• This is a joined effort from @dmigo and me. We implemented the reciprocal rank fusion method to merge results coming from different retrievers. The method is described here: https://plg.uwaterloo.ca/~gvcormac/cormacksigir09-rrf.pdf

Status (please check what you already did):

• First draft (up for discussions & feedback)
• Final code
• Added tests
• Updated documentation

[ZanSara]

Looks good but there are a few possible improvements 🙂 Please don't hesitate to question my comments if they don't feel right! I tend to be a bit tedious sometimes 😄

[mathislucka]

Hey @ZanSara thanks for the thorough review :) @dmigo and I refactored a little bit and it's hopefully a cleaner version now.

[ZanSara]

Nice, it's good to go for me!

If you have the time, though, I'd split test_join_document_pipeline into several smaller tests. Why having such a huge test when the units within it are so visible? 🙂

[ZanSara]

One last comment: please run mypy locally before merging, as I noticed a bug into one of my latest CI changes that sometimes make the check fail for a cache miss. If mypy shows no errors, then it's good to merge: the Ray issue is known and will not cause problems.

[dmigo]

If you have the time, though, I'd split test_join_document_pipeline into several smaller tests.

😆 we've also discussed that that would be nice to do