Allow a list of keys in the `by` option of the v2 version of the `join` operator

[nvnieuwk]

New feature

Currently the by option in the .join operator for typed workflows allows only a single key to be specified. This on it's own is a very useful feature and will greatly improve how we write pipelines. It would however also be helpful if this option could take a list of multiple keys and only join entries where all of these keys match. This would prevent pipeline developers from having to do a .map on both channels before joining them. It would also increase the readability of the pipeline as it's clear from the join operator what keys are used exactly (instead of one custom key created just for joining)

Use case

Joining channels with records that have multiple keys in common

Suggested implementation

ch1.join(ch2, by['id', 'type']) or similar

[itrujnara]

From my experience, I would also suggest allowing a closure for the by argument to make the code cleaner when the keys have different names on each side. Example:

ch_samples.join(ch_genomes, by: { sample, genome -> sample.genome == genome.id })

This might be a slightly niche situation (since usually the joined channels have the same shape), but I think it could be handy in some real scenarios.

[bentsherman]

Yep, let's just skip the list of keys and go straight to closure. That's also how a SQL join works (the WHERE clause)

[nvnieuwk]

Sounds good! Thank you!