Grim reaper

[wneirynck]

Even though there are plenty of checks, it may occur that a container instance keeps on running forever. This may happen for example when the process that is checking the termination (script or api) is killed or restarted. For this, we should have a "grim reaper" that checks if there are any dangling containers that need to be killed off. This could for example be run as a kubernetes cronjob. It should check all container instances that have been running longer than the configured timeout. It could possible mark the related builds as ended.