release-21.2: opt: fix hoist of ANY comparison with tuples

[mgartner]

Backport 1/1 commits from #98700.

/cc @cockroachdb/release

---

opt: fix hoist of ANY comparison with tuples

Prior to this commit, when hoisting Any expressions like
<left> = ANY (SELECT <right> ...), we constructed
(IsNot <left|right> Null) expressions which are equivalent to
<left|right> IS DISTINCT FROM NULL. As discovered in #46675, these
expressions have different behavior than <left> IS NOT NULL when
<left> is a tuple. As a result, the hoisting transformations could
construct invalid plans that cause incorrect results. This commit fixes
the issue by using IsTupleNotNull expressions when <left> and
` are tupleq.

Fixes #98691

Release note (bug fix): A bug has been fixes that caused incorrect
results of ANY comparisons of tuples. For example, an expression like
(x, y) = ANY (SELECT a, b FROM t WHERE ...) could return true
instead of the correct result of NULL when x and y were NULL, or
a and b were NULL. This could only occur if the subquery was
correlated, i.e., it references columns from the outer part of the
query. This bug was present since the cost-based optimizer was
introduced in version 2.1.

Release justification: Fixes a minor bug.

[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 

[yuzefovich]

Do we need 21.2 backport? I thought we only keep doing security backports there.

[mgartner]

Oh no. I can't keep all the 1's and 2's straight in the last 3 versions. Excited for introducing a third digit in 23.1!