When the queue is down or non-responsive, we end up with long exception

traces in the log files but nothing useful on the front-end. Tool

updates and tool runs will spin and spin until they time out.
We now have error handling code around all code communicating over

Celery, logging errors and stack traces with a trace ID.
Tool runs (both manual and automatic) will fail to the error state with

an "error running tool (error ID XYZ)", instead of waiting to time out.
Worker status checks will once again display a suitable error message

and include information in the logs. We had code that attempted to

provide a good result, but it was checking for IOError, and nothing in

the Celery communication code uses that anymore.
Tool refreshes in the Tools database list now alert with an error and

reset the state. This has also been updated to be more accessible and

prioritize Ink styling if available.

Shut down RabbitMQ and tested all the functionality: Manual tool runs,

automatic tool runs, worker status checks, and tool refreshes. Verified

the log output and the user-facing output in all cases.
Tested the newer tool refresh UI on Review Board 6 and 7.
Unit tests pass.