sql: non-deterministic allocation of desc IDs after a 3-node TestCluster initialization

[knz]

Describe the problem

Found in #110242: after the TestDataDriven initializes a 3-node cluster (and NOT in the 1-node case), the test sometime fails while expecting the initial table to be generated with ID N, and sees instead ID N+1 or N+2.

This only occurs in CI -- cannot be reproduced locally under stress.

We can work around the issue by normalizing the expected table ID in the test (to become agnostic of the numeric value) however the fact that there is unstable ID allocaiton in a 3-node TestCluster is abnormal and unexpected.

To Reproduce

Take the code from #110492
Remove the ID normalization that was added in that PR.
Then exercise the test in CI (presumably it does not fail when run locally)

Expected behavior

We want to understand what is the source of the non-determinism here.

(NB: this issue is not about fixing the situation for the tests. The proper longer-term fix will be a combination of #110491 and ensure that all test expectations use names from that change.)

Jira issue: CRDB-31461

[rafiss]

I think we figured this one out - @ecwall fixed this by making the test use the TransactionalDescIDGenerator #110829