Feature: keep jobs around after successful completion

[ghost]

Jobs that have completed successfully are removed once the milestone advances. This makes it hard to verify that the execution really was successful, or what exactly the job executed.

Only delete jobs no longer in the target state when they're older than a configurable amount of time, for example, 15 minutes. This keeps the pod(s) around as well, and allows easy inspection of the logs.