cdc: surface repetitive or likely permanent errors

[shermanCRL]

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

Changefeeds have switched to “default retryable” on error. This is good!

We have a new problem, perhaps, that errors will retry a lot or indefinitely. This likely means the error is de facto permanent -- maybe a config problem, auth, etc.

Describe the solution you'd like

These situations should be surfaced to the user.

I don’t know how! The starting place would be to define “too repetitive” or “probably permanent”. The surfacing could be a metric? An alert in DB Console?

Jira issue: CRDB-28643

[blathers-crl]

cc @cockroachdb/cdc

[amruss]

Closing this, its too vague. We do have: #103539