Add a healthcheck to Supervisor and have a listener trigger a rebuild if it becomes unhealthy

### Description

If Supervisor had a healthcheck in its image then a listener could trigger a rebuild if it became unhealthy automatically. This would help users in situations like #6331 (and the other issues linked from that one) for example without resorting to drastic measures like a reinstall or using the host shell to manually remove the container.

Observer and OS Agent both have access to the docker socket and so could be the listener here that triggers the rebuild. OS Agent also has sentry access I believe so it could possibly report any info it could collect on Supervisor first to help us debug how Supervisor got into a stalled state in the first place and stop that from happening. Though the particular linked issue doesn't seem to have any useful information available unfortunately so I'm not sure how helpful this would be.

### Additional context

_No response_

Start by reviewing how the Supervisor image currently runs and how Observer or OS Agent accesses the Docker socket and handles rebuilds. No file or test is identified in the issue; the work is complete when an unhealthy Supervisor is detected, a listener safely triggers its rebuild, and the relevant diagnostic behavior is agreed and tested.

Written by the indexing model from the issue text.