backup,restore: detect disallowed/failing implicit credentials on job creation

**Is your feature request related to a problem? Please describe.**

We are seeing a class of error in the logs where a user has specified implicit auth for their backups, the implicit auth is disabled on that cluster (via launch flag), and thus the backups are failing.

It seems like a better UX that they should have gotten the error upon creation, in situations where CRDB can know that it will not succeed.

**Describe the solution you'd like**

Do some light validation of implicit credentials upon creation of a backup/restore/schedule, and respond via the SQL session.

A simple, narrow version can simply check that implicit auth is enabled at all -- it wouldn’t require I/O. Or, we could ping as we do with external connections (with short timeouts, presumably).

**Describe alternatives you've considered**

We’ve generally avoided I/O during job creation, as that expense is more appropriate for job execution. The above could presumably be done outside of a transaction and kept lightweight.

Jira issue: CRDB-29139