job-runner: use volumes for artifacts

[jelly]

Our job-runner has issues cleaning up containers mostly stemming from not running with podman run --rm but creating a cidfile, and then cleaning it up later.

The reason we use a cidfile is that our jobs work as following

• create a container and dump a cidfile
• once tests are done, copy the logs out of the container using the cid
• podman rm with the cid

The podman rm can fail sometimes so as an alternative we could bind a temporary directory and dump the logs there. The temporary directory would be created and maintained by job-runner (and removed on exit)

[jelly]

One concern is that I have seen a job container deadlock that would not be resolved by this and would require us to implement a proper timeout mechanism.

[jelly]

One thing I forgot to mention is that we run our job-runner in a container so we can't just create a temporary directory and share it with the spawned container as that would be spawned on the host (next to the job-runner container).

So we would need to extend ansible/roles/tasks-systemd/tasks/main.yml to pass a volume like cockpituous-artifacts to store the temporary image results.