Option to only sample every nth core, and skip the rest.

[damster101]

Would be useful for things like skipping measurements between the same physical core due to SMT, or focusing purely on latencies between even larger subgroups, like AMD CCXs.