sql: fix another buglet with tenant settings

[knz]

This (hidden) bug was left over from #108902.
There is no visible negative effect to this bug being present until setting aliases are effectively introduced, which hasn't happened yet.

The code modified here is already properly exercised by logic tests.
Also, this is exercised by #109077 and #109074 and the CI in those PRs readily fail without this fix.

Epic: CRDB-27642

[blathers-crl]

It looks like your PR touches production code but doesn't add or edit any test code. Did you consider adding tests to your PR?

_{🦉 Hoot! I am a Blathers, a bot for CockroachDB. My owner is dev-inf.}

[cockroach-teamcity]

This change is 

[stevendanna]

If I'm reading this correctly, the bug here is that the crdb_internal.cluster_settings table was returning the name which was then being used in a join against the underlying systems table where the join actually has to happen on the setting_key.

If we wanted to add a test in this PR, I suppose we could make a test that registers a new setting that won't be in production builds and then write the test against that.

But, given that you seem to be iterating through test failures from the real-world case, we do know we have test coverage so perhaps we don't need to duplicate it.

[knz]

If I'm reading this correctly, the bug here is that the crdb_internal.cluster_settings table was returning the name which was then being used in a join against the underlying systems table where the join actually has to happen on the setting_key.

Yes.

given that you seem to be iterating through test failures from the real-world case, we do know we have test coverage so perhaps we don't need to duplicate it.

Correct

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

[knz]

TFYR!

bors r=stevendanna

[craig]

Build succeeded:

• Bazel Essential CI (Cockroach)