kv: permit some non-transactional batches to perform follower reads

[nvanbenschoten]

Touches #67562.

This commit adds support for a subset of non-transactional batch requests to perform follower reads. Specifically, it makes those that do not rely on their timestamp being set from the server's clock eligible.

This condition is necessary because if a follower with a lagging clock sets its timestamp and this then allows the follower to evaluate the batch as a follower read, then the batch might miss past writes served at higher timestamps on the leaseholder.

[cockroach-teamcity]

This change is 

[aayushshah15]

I'm missing some context here, particularly around when we would/wouldn't expect a non-transactional request to receive its timestamp from the server-side. Is there anything you could point me to regarding this? Otherwise, I can bring it up offline with you later.

Reviewable status: complete! 1 of 0 LGTMs obtained

[nvanbenschoten]

TFTR!

bors r+

Is there anything you could point me to regarding this? Otherwise, I can bring it up offline with you later.

I think the best resources about this are BatchRequest.SetActiveTimestamp and Header.Timestamp, but neither of these is comprehensive. This issue about removing them might actually be more helpful: #58459. The way to think about this is that allowing a non-transactional request (which is always single-range) to get its read timestamp from the server is how we can ensure single-key linearizability without uncertainty intervals. But there is a class of non-transactional requests that aren't concerned with linearizability and are fine with staleness. Instead, they have a specific timestamp that they want to operate at. These are the requests that set a timestamp on the client and these are the requests that this PR permits to run on suitable followers (subject to other requirements in BatchCanBeEvaluatedOnFollower).

[craig]

Build succeeded:

• GitHub CI (Cockroach)