feat(shell-api): use collStats aggregation stage instead of command for collection.stats() MONGOSH-1157

[Anemy]

MONGOSH-1157

This PR updates our db.collection.stats() shell api method to use the aggregation stage $collStats for fetching collection statistics instead of the db.runCommand({ collStats: collectionName }) form. The run command form is to be deprecated by 7.0 (PM-2362)

The $collStats stage returns a document for each individual shard's statistics. A lot of the code of this pr involves aggregating those documents into one document to be returned. I initially was trying to do this all in one aggregation pipeline, however I found it was going to be a bit harder to test and maintain. Things like new keys like numOrphanDocs added in 6.0 make an aggregation a lot more complex. If curious a fairly non-optimized implementation (without timeseries support) can be found on this gist: https://gist.github.com/Anemy/1e3b38f8f4f07f055a8c18d4b6ab443b

The $collStats aggregation stage fails on sharded timeseries collections so we fallback to the old command form when that error code is encountered.

Notes

• These changes alter the output format of db.collection.stats()
  Using the aggregation $collStats stage does not return the same data
  as the soon to be deprecated collStats command. Namely:
  - primary is not given now.
  - nchunks is not given now.
  These fields already aren't listed in our documentation: https://www.mongodb.com/docs/manual/reference/command/collStats/ but I'm curious to hear if folks feel we should work to re-add them or have more insights.

[addaleax]

Is the way we're checking for sharding okay? Is there a better way? Does the current way require special permissions?

Yeah, I think it does require permissions that not all users might have… we could maybe fall back to doing something like the (somewhat inaccurate) collStats.length > 1 if we cannot query config.collections?