[k8s] Define semantic conventions for k8s cpu metrics

[ChrsMark]

Area(s)

area:k8s

Is your change request related to a problem? Please describe.

As part of the K8s SemConv stability work we need to define k8s cpu metrics that are already in use by the Opentelemetry Collector.

Related to #1032

Describe the solution you'd like

The existing metrics in use that we need to define as semantic conventions are the following:

k8s.node.cpu.usage:✅ #1320
k8s.node.cpu.utilization: Blocked until open-telemetry/opentelemetry-collector-contrib#27885 (comment) is covered
k8s.node.cpu.time: ✅ #1320
k8s.pod.cpu.usage: ✅ #1320
k8s.pod.cpu.utilization: Deprecated
k8s.pod.cpu.time: ✅ #1320
container.cpu.usage: ✅ #1128
container.cpu.utilization: Deprecated
container.cpu.time ✅ #282
k8s.container.cpu_limit_utilization
k8s.container.cpu_request_utilization
k8s.container.cpu.node.utilization
k8s.pod.cpu.node.utilization
k8s.pod.cpu_limit_utilization
k8s.pod.cpu_request_utilization

Describe alternatives you've considered

No response

Additional context

No response

[ChrsMark]

I wonder if we could/should change *.cpu_limit_utilization to *.cpu.limit.utilization and *.cpu_request_utilization to *.cpu.request.utilization.

Along with the following (from #1032):
k8s.container.cpu_request -> k8s.container.cpu.request
k8s.container.cpu_limit -> k8s.container.cpu.limit

I know this would be a breaking change in the Collector but I want to check if we see value in doing this for consistency.
Also @lmolkova I can recall you had an issue with dots VS underscores where dots were preferred over underscores. Do I remember correctly here? Is this still valid?

[ChrsMark]

Correcting myself after discussing this in the K8s SemConv stability SIG:

We cannot have k8s.container.cpu.request and k8s.container.cpu.request.utilization because of the constraint that metrics cannot be namespaces at the same time.

I don't have a better alternative here than what we have already defined in the Collector but if anybody has feel free to share it.

[ChrsMark]

Correcting myself after discussing this in the K8s SemConv stability SIG:

We cannot have k8s.container.cpu.request and k8s.container.cpu.request.utilization because of the constraint that metrics cannot be namespaces at the same time.

I don't have a better alternative here than what we have already defined in the Collector but if anybody has feel free to share it.

Thinking of it again I think we could go with sth like k8s.container.cpu.request (for #2074) and k8s.container.cpu.request_utilization. I believe having the cpu. as a common namespace for all metrics would be ideal. @open-telemetry/semconv-k8s-approvers @Roger wdyt?