worker: raise hell when failing to join the run channel

When attempting to join the run channel without a valid WORKER_RUNS_PRIVATE_KEY set, the worker will silently fail to make claims.

Frustratingly we don't get any kind of error details from lightning, and the actual error event that comes through Phoenix is erratic, so say the list. I have to jump through hoops to find an error object, and when I get it there's no context - just an empty object.

See community thread

I think we need to things here:

1. First Lightning needs to properly reject the token. I don't know why it doesn't right now.
2. Second Lightning probably needs to return the run to the queue or something
3. We need to work out what to do in this case. Probably we need to kill the worker with a clear error message.

Users encountering this in Docket at the moment are getting zero feedback . Even if there's an error returned, I'm not sure it'll be very clear in the logging.

We could consider:

1. The worker should very noisily error when there's no token detected in the env. Just process.exit on startup#
2. ?

The issue names the worker startup path, run-channel join, WORKER_RUNS_PRIVATE_KEY, and Lightning/Phoenix error handling but no files or tests. Start by locating those entry points and tracing missing credentials through startup and claim handling. Done requires an agreed behavior, a clear surfaced error, and an explicit outcome for the failed run.