release-23.2: rangefeed: cancel async tasks on processor stop

[erikgrinaker]

This is part of a #117612 backport.

This issue only affects the new scheduler-based rangefeed processor in 23.2 (off by default), the legacy processor handled this properly.

---

Backport 2/2 commits from #117859.

Release justification: prerequisite for backporting #117612.

/cc @cockroachdb/release

---

Previously, async tasks spawned by the rangefeed processor (typically txn pushes and resolved timestamp scans) were not cancelled when the processor was stopped or the stopper quiesced. If these operations stalled, this could lead to goroutine leaks and node shutdown stalls. However, this was mitigated to some extent by the intent resolver itself detecting stopper quiescence.

This patch uses a separate task context for async tasks, which is cancelled either when the processor is stopped or the stopper quiesces.

In general, the rangefeed scheduler shutdown logic could use some improvement, but this patch does not attempt a broader cleanup in the interest of backportability.

Epic: none
Release note: None

[blathers-crl]

Thanks for opening a backport.

Please check the backport criteria before merging:

• Backports should only be created for serious
  issues or test-only changes.
• Backports should not break backwards-compatibility.
• Backports should change as little code as possible.
• Backports should not change on-disk formats or node communication protocols.
• Backports should not add new functionality (except as defined
  here).
• Backports must not add, edit, or otherwise modify cluster versions; or add version gates.
• All backports must be reviewed by the owning areas TL and one additional
  TL. For more information as to how that review should be conducted, please consult the backport
  policy.

If your backport adds new functionality, please ensure that the following additional criteria are satisfied:

• There is a high priority need for the functionality that cannot wait until the next release and is difficult to address in another way.
• The new functionality is additive-only and only runs for clusters which have specifically “opted in” to it (e.g. by a cluster setting).
• New code is protected by a conditional check that is trivial to verify and ensures that it only runs for opt-in clusters. State changes must be further protected such that nodes running old binaries will not be negatively impacted by the new state (with a mixed version test added).
• The PM and TL on the team that owns the changed code have signed off that the change obeys the above rules.
• Your backport must be accompanied by a post to the appropriate Slack
  channel (#db-backports-point-releases or #db-backports-XX-X-release) for awareness and discussion.

Also, please add a brief release justification to the body of your PR to justify this
backport.

[cockroach-teamcity]

This change is 

[rharding6373]

Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @nvanbenschoten and @rharding6373)

---

pkg/kv/kvserver/rangefeed/task.go line 64 at r2 (raw file):

	if err := s.iterateAndConsume(ctx); err != nil {
		err = errors.Wrap(err, "initial resolved timestamp scan failed")
		if ctx.Err() == nil { // cancellation probably caused the error

Out of curiosity, why do we not want to log these errors when it wasn't caused by a cancellation?

[erikgrinaker]

Out of curiosity, why do we not want to log these errors when it wasn't caused by a cancellation?

We log the error when it wasn't caused by cancellation. If it's cancelled, the system is shutting down, so these are just noise -- potentially hundreds of them.