Option for on-request debugging of a consumer group

Nobody has claimed this yet.

While debugging a problem with ZK offset checking, I found it was necessary to have some in-depth debugging of a particular consumer group so I could watch it. I added debug logging for offset commits to the code, but this can be overwhelming to turn on for a busy instance. It would be much nicer to turn on debug logging for a particular group in a particular cluster with a request to the API, and disable it again when done.

This shouldn't be too difficult. We just need to maintain a list of (cluster, group) tuples to emit debug logging for, similar to the group blacklist. It could even be just group (over all clusters) to simplify things. Probably makes sense to add a common debug logging method that encapsulates this logic.

Start by reading the existing group blacklist and offset-commit debug logging paths, then trace the API entry points for enabling and disabling diagnostics. Done means a request can toggle debug logging for a selected consumer group, with the scope and behavior matching the issue.

Written by the indexing model from the issue text.