Add support for CPU and Wall time for async-profiler

[ambud]

We need the ability for async-profiler to be run in cpu time spent vs wall time spent allowing developers to see what is computationally expensive vs what is blocking.

For starters can we implement it for dynamic profiling only to have a quick implementation for this? if that works well then we can find a way to export it to continuous profiling.

For continuous profiling implementation my idea is to leverage fleet specific configuration so we are recording what the owner of the fleet is most interested in monitoring (cpu vs wall). However, perf engineers would want both capabilities and that can be satisfied using dynamic profiling.