fix: recreate version probe Job after pod GC

[vitaliytv]

Completed version probe Jobs can outlive their Pods after Pod garbage collection. The controller then repeatedly fails to read the result and blocks reconciliation with no pods found for version probe job.

When a completed Job has no Pods, delete the Job and let the next reconcile create a fresh probe. Other read failures keep their existing behavior.

Validation:

• go test ./internal/controller -count=1 -ginkgo.v
• make lint

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}