sql: missing predicates to start the job scheduler in secondary tenants

[knz]

Describe the problem

The job scheduler is an async, tenant-scoped SQL-level process that scans system.jobs periodically.

To avoid contention on the system table, we have a mechanism intended to only start the job scheduler on 1 node of the cluster.
(The ShouldRunScheduler on JobExecutionConfig)

Today, the mechanism only work on the system tenant.

It does not work on secondary tenants.

This means that any use of secondary tenants (starting with CC serverless, and will extend to UA clusters) will experience job contention just as when the mechanism didn't exist yet.

Expected behavior

We need to find a better condition that works for all tenants, not just the system tenant.

Epic: CRDB-14537

Jira issue: CRDB-20745