Is your feature request related to a problem? Please describe.
Currently, if a fluent-server log sink in CRDB fails to connect to the fluent-server target, a log message is emitted indicating the failed connection attempt.

It would be nice to also have this tracked in the form of a counter metric, which can help us more easily correlate this failure with other metrics, such as those emitted by fluent-bit processes.

Describe the solution you'd like
Add a counter metric that's incremented each time a fluent-server log sink fails to connect to the target. Make sure that fluent-server log sinks with buffering enabled (wrapped in a bufferSink) still increment these metrics on connection failure.

Jira issue: CRDB-27635

Epic CC-9681