Tracker for https://github.com/kubevirt/kubevirt/issues/11731

What happened:
The downward metric server could be abused to use an excessive amount of resources by the virt-handler if many VMs are running on the cluster and requesting the metrics.

This was first brought up on CNV-38661 and mitigated by #11484, but even with this mitigation in place, the root issue still persists: the server runs in the virt-handler outside the VM's cgroup.

A proper solution will be to move back the downward metrics server from the virt-handler to the virt-launcher, so it will run in the same cgroup as the VM