ui: do not retry table metadata polling requests when a 5xx is returned

[dhartunian]

We should add a rate limit for the retries to the api/v2/table_metadata/updatejob/?onlyIfStale=true" call and use exponential decay. Current behavior will spam the cluster and frontend logs when a rolling restart is happening

Jira issue: CRDB-50499

[blathers-crl]

Hi @dhartunian, please add branch-* labels to identify which branch(es) this C-bug affects.

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