Feature request: Add a bitmask to enable individual perf context counter

Currently we can change perf level to enable different set of perf context counters. However, sometimes we are only interested in a subset of timer counter and want to avoid overhead of the other counters, and using only perf level doesn't fit out needs. We propose allowing user to set a bitmask which can enable individual perf context counters. Currently there are ~70 perf context counters, so we only need an uint64[2] array to hold the bitmask.

We would like to get some feedback on whether this is a good idea, and can work on a PR later.