Fix ProxmoxMachine deletion to wait for stop/destroy tasks

[Leonorus]

Issue #, if available:
Fixes #742

Description of changes:
- Make ProxmoxMachine deletion task-aware for async Proxmox qmstop and qmdestroy tasks.
- Return and track the qmstop task for running VMs instead of immediately submitting qmdestroy while the VM may still be running.
- Store deletion task UPIDs in status.taskRef and requeue while stop/destroy tasks are still running.
- Keep the finalizer and node location until deletion is confirmed by a successful destroy task and VMID-free check.
- Surface deletion task failures with VirtualMachineProvisioned=False, Reason=DeletionFailed, and the Proxmox task exit status, e.g. qmdestroy: VM 103 is running - destroy failed.
- Preserve safe cleanup paths for already-free VMIDs, missing/expired tasks, and unassigned/reserved VMIDs.
- Add regression coverage for deletion task tracking, failed destroy handling, VMID-free confirmation, transient task lookup failures, and running-VM stop-before-destroy behavior.

[wikkyk]

Don't worry about the test 'failure' - it's only the sonarqube report step that's failing (I'm working on improving the workflow).

[wikkyk]

Thanks. I'll let @65278 do another review because he knows that part of the code better.