feature request: make node cache metrics listen port reusable

[paveew]

Since kubernets1.25 rollouts maxSurge for DaemonSets, it's easier to support zero-downtime upgrade for node-local-dns.

If we set maxSurge to 1 and maxUnavailable to 0 for DaemonSet, a new pod will created before the old one is deleted during an upgrade, which is an expected behavior and makes node-local-dns always available.

However, the new node-local-dns pod attempts to listen port 9353 for metrics, which produces an address already in use error because the port is currently in use by the old node-local-dns pod.

If node-local-dns supports reuse of ports, i.e. SO_REUSEPORT, it would be great for this case.

node-local-dns HA reference：#324