Add metrics for cpumanager

[ffromani]

What would you like to be added?

Add metrics to enable observability of the cpumanager.

Why is this needed?

This is required to GA the feature but it's also useful in general.

To the best of my knowledge, it's very hard to observe the behavior of cpumanager without peeking into the pod and/or use the podresources interface (kubelet >= 1.24). But even for the latter, no well known components expose metrics consuming this information.

I believe anyway kubelet some expose some basic metrics about cpumanager behaviour, while deferring more fine grained metrics to external components (NFD, node_exporter, is fine)

[ffromani]

/sig node
/assign

[swatisehgal]

/triage accepted
/priority important-soon
(as this is important for graduation for CPU Manager to GA)