Improve error handling in the cleanup artifact worker concern

🔥 Problem

In the package registry and the dependency proxy, we have this requirement of go through a set of records marked as destroyed and actually destroy them. Workers there are quite simple: select one marked record and destroy it.

This action is so simple and similar in the package registry and the dependency proxy that we created a concern to share common logic.

Related to !84700 (merged), we saw several flaws in that concern:

1. There is status updates going on.
   1. this already brought issues (#357373 (closed))
   2. we don't see the value of such status updates. Observability is already good enough without it.
   3. we currently don't have any handling of objects that stays in the processing or error status.
2. Errors are swallowed.
   1. This makes hard to debug issues as the original error is hidden and will not appear in the logs.

🚒 Solution

1. Remove all status updates.
   • In the end, what matters is: either the worker destroyed the record or not. If not, then we need to log it (already done by the fail message of the worker and the considered record needs to stay as it is so that it is retried.
2. Remove the rescue block.
   • Don't try to handle errors and simply let them bubble up.
   • This will create a fail message that will log the error and its backtrace = much smoother debugging experience.