CPU statistics in containers limited via CFS are misleading

There are a couple of different ways to limit the CPU that can be used by a container.
A common way used by Docker is specify --cpuset-cpus, which limit the container to only be able to use specific cores within the CPU.
This is currently auto-detected by ns_server and memcached (MB-28898. MB-20834).

However, Kubernetes limits CPU for containers using CFS instead (reference: https://medium.com/@betz.mark/understanding-resource-limits-in-kubernetes-cpu-time-9eff74d3161b).

Couchbase Server does not account for this, so will incorrectly report the number of CPUs that a Kubernetes pod is using and sigar will also report an incorrect total CPU utilization rate.

For example, in a container limited to a single CPU core:

"cpuCount":2

$ cat cpu.cfs_period_us

100000

$ cat cpu.cfs_quota_us

100000

I expect that this is not easy to fix, as we'd have to use the CFS quotas and share to calculate the number of CPUs, but raising this for awareness.

I also suspect that this may also affect the automatic thread generation used by memcached too, but have not verified this.