release-23.2.6-rc: sql/stats: evict stats cache entry if user-defined types have changed

[michae2]

Backport 1/3 commits from #124603.

/cc @cockroachdb/release

---

sql/stats: evict stats cache entry if user-defined types have changed

When adding table statistics to the stats cache, we decode histogram upper bounds into datums. If the histogram column uses a user-defined type, we hydrate the type and use this to decode.

In statistics builder, these histogram upper bound datums are compared against datums in spans and constraints. The comparisons assume that the datums are of equivalent type, but if the user-defined type has changed sometime after loading the stats cache entry, this might not be true.

If the user-defined type has changed, we need to evict and re-load the stats cache entry so that we decode histogram datums with a freshly-hydrated type.

(We were already checking UDT versions when building the optTable in sql.(*optCatalog).dataSourceForTable, but the newly-built optTable used the existing table statistics instead of refreshing those, too.)

Fixes: #124181

Release note (bug fix): Fix a bug where a change to a user-defined type could cause queries against tables using that type to fail with an error message like:

histogram.go:694: span must be fully contained in the bucket

The change to the user-defined type could come directly from an ALTER TYPE statement, or indirectly from an ALTER DATABASE ADD REGION or DROP REGION statement (which implicitly change the crdb_internal_region type).

This bug has existed since UDTs were introduced in v20.2.

---

Release justification: fix for a serious production issue.

[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 

[michae2]

I looked in depth at the Bazel Extended CI timeouts, and they seem to be the normal timeouts for stress-race, experienced frequently even before this PR. There is no deadlock like I feared.

[mgartner]

Nicely done!

Reviewed 11 of 11 files at r1, all commit messages.
Reviewable status: complete! 1 of 0 LGTMs obtained (waiting on @michae2 and @rafiss)

---

pkg/sql/stats/stats_cache.go line 320 at r1 (raw file):

// getTableStatsFromCache is like GetTableStats but assumes that the table ID
// is safe to fetch statistics for: non-system, non-virtual, non-view, etc.

nit for follow-up PR: explain the udtCols argument.