release-2.1: storage: take an engine checkpoint during failing consistency checks

[irfansharif]

Backport 1/2 commits from #36867.

Useful as part of #42011.

+cc @cockroachdb/release

---

This takes a checkpoint on the nodes with replicas of a failing range,
before the failure leads to nodes shutting down. The checkpoint will, for
the replicas of the affected range, be taken at the same Raft log position.

Release note: None

[cockroach-teamcity]

This change is 

[tbg]

Clean backport? Wonderful!

[irfansharif]

Failed with #31778, fix in #32899 was not backported. Retrying.

[irfansharif]

Failing with Example-ORM failures with what looks like network flakes?

[TestDjango/FirstRun] main_test.go:163: Get http://localhost:6543/ping/: dial tcp 127.0.0.1:6543: getsockopt: connection refused

If safe to ignore, how do I get past the failing TC check here?

[irfansharif]

Ok, I see the same example-orms failures on release-v2.1 (was a major PITA to get this running, so much is broken). Very surprisingly the example-orms test suite isn't a pinned version, the changes made on example-orms master are reflected in test suites run against all branches of CRDB thereonforth. This is almost definitely not what we want.

@tbg: It's safe to disable branch protection for this PR, I don't have access to do so.

+cc @rohany, @rafiss: all future backports to release-v2.1 (and likely others) are going to fail. I don't know if this is something we care about, but TC will fail+block for future backports.

[tbg]

Merging at your request