release-24.2: stats: use no-op DatumAlloc when decoding EncDatums

[blathers-crl]

Backport 2/2 commits from #136780 on behalf of @yuzefovich.

/cc @cockroachdb/release

---

tree: make nil DatumAlloc act as a no-op

This commit makes so that nil DatumAlloc value makes the alloc struct a no-op, i.e. each datum type passed by value results in a separate allocation. This can be useful when the callers will keep only a subset of all allocated datums.

Release note: None

stats: use no-op DatumAlloc when decoding EncDatums

This commit fixes a bounded memory leak that could previously occur due to usage of the DatumAlloc when decoding EncDatums into tree.Datums during table stats collection. We have two use cases for decoded tree.Datums:

• we need it when adding all rows into the sketch (we compute the datum's fingerprint ("hash") to use in the distinct count estimation). This usage is very brief and we don't need the decoded datum afterwards.
• we also need it when sampling some rows when we decide to keep a particular row. In this case, the datum is needed throughout the whole stats collection job (or until it's replaced by another sample) for the histogram bucket computation.

The main observation is that the DatumAlloc allocates datums in batches of 16 objects, and even if at least one of the objects is kept by the sample, then all others from the same batch are as well. We only perform memory accounting for the ones we explicitly keep, yet others would be considered live by the Go runtime, resulting in a bounded memory leak. This behavior has been present since forever, but it became more problematic in 24.2 release with the introduction of dynamically computing the sample size. To go around this problematic behavior this commit uses nil DatumAlloc which makes it so that every decoded datum incurs a separate allocation. This will have a minor performance hit and increase in total number of allocations, but at least most of them should be short-lived. Alternatively, we could use an "operating" DatumAlloc during fingerprinting and a no-op during the sampling, but we'd need to explicitly nil out the decoded datum after fingerprinting which would result in decoding some datums twice.

Fixes: #136394.

Release note: None

---

Release justification: bug fix.

[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. 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