descmetadata: do not use experimental setting

[yuzefovich]

Release justification: low-risk cleanup.

Release note: None

[cockroach-teamcity]

This change is 

[yuzefovich]

I don't quite understand why this was introduced in #83592 - am I missing something? The CI seems happy.

[fqazi]

The concern was if ExperimentalDistSQLPlanningAlways was somehow set inside the session, anything inheriting the session run with that. That could cause internal inserts from the metadata updater, etc.. to fail. This used to be a problem on logictest at one point

[yuzefovich]

I see. This would only be applicable to a couple of logic test files (that explicitly do SET experimental_distsql_planning = always), so I think it's not good that we create a copy of the session data in all cases, production including. I briefly stressed the relevant files with this PR and didn't see any flakes, so I think we should proceed with this change. If the flakes do happen, then we probably will need to add RESET experimental_distsql_planning to the end of those files.

[fqazi]

Reviewed 2 of 2 files at r1, all commit messages.
Reviewable status: complete! 0 of 0 LGTMs obtained (waiting on @ajwerner)

[yuzefovich]

TFTR!

bors r+

[fqazi]

I'm okay with that approach, the onus should be on the tests themselves. We used to have a logictest mode that forced it in other cases I think

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

[yuzefovich]

We did have -spec-planning configs (which were removed like a month ago), but they were using On and not Always for this experimental setting.

[craig]

Build succeeded:

• Bazel Essential CI (Cockroach)