release-22.1: colexec: fix incorrect accounting when resetting datum-backed vectors

[yuzefovich]

Backport 1/1 commits from #97750.

/cc @cockroachdb/release

---

This commit reverts a couple of other commits:

• "colexec: fix a "fake" memory accounting leak for intra-query period"
  (72e83fe)
• "colexec: deeply reset datum-backed vectors in ResetInternalBatch"
  (cb93c30)

since they introduced incorrect memory accounting for the datum-backed
vectors.

Those two commits together solved another issue where we would keep
no-longer-needed datums live for longer than necessary (until they are
overwritten in the datum-backed vector) by eagerly nil-ing them out when
resetting the whole batch. This required introducing some careful
adjustment to the memory accounting in order to keep the accounting up
to date. However, that logic turned out to be faulty; in particular, it
became possible to register the allocations of the datum-backed vectors
with one account but then attempt to release some of those allocations
from another. If those releases happen enough times, it'd put the
account in debt which would trigger an internal error (or a crash in
test builds).

Such a scenario can occur because we have a couple of utility operators
that append a vector to a batch owned by another operator. When that
other operator resets its batch, the appended-by-utility-operator
vector is also reset, and the memory usage of the freed datum would be
deregistered from the wrong account. Tracking precisely which vector is
owned by the owner of the batch vs appended by another operator can be
cumbersome and error-prone, so this commit instead of introducing this
tracking removes the resetting behavior of the datum-backed vectors.
This should be bullet-proof while only increasing slightly the amount of
time references to datums are kept live.

Fixes: #97603.

Release note (bug fix): CockroachDB could previously encounter an
internal error "no bytes in account to release ..." in rare cases and
this is now fixed. The bug was introduced in 22.1.

Release justification: bug fix.

[blathers-crl]

Thanks for opening a backport.

Please check the backport criteria before merging:

• Patches should only be created for serious issues or test-only changes.
• Patches should not break backwards-compatibility.
• Patches should change as little code as possible.
• Patches should not change on-disk formats or node communication protocols.
• Patches should not add new functionality.
• Patches must not add, edit, or otherwise modify cluster versions; or add version gates.

If some of the basic criteria cannot be satisfied, ensure that the exceptional criteria are satisfied within.

• 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.
• The PM and TL on the team that owns the changed code have signed off that the change obeys the above rules.

Add a brief release justification to the body of your PR to justify this backport.

Some other things to consider:

• What did we do to ensure that a user that doesn’t know & care about this backport, has no idea that it happened?
• Will this work in a cluster of mixed patch versions? Did we test that?
• If a user upgrades a patch version, uses this feature, and then downgrades, what happens?

[cockroach-teamcity]

This change is 

[rharding6373]

Reviewable status: complete! 1 of 0 LGTMs obtained (waiting on @DrewKimball)