Record instruction count across all threads created by the benched code

[rukai]

I think it should be possible because rustc is multithreaded and measures instruction counts for its benchmarks https://perf.rust-lang.org/?start=&end=&kind=raw&stat=instructions%3Au

[jbreitbart]

I have not checked this, but it should already be possible. Can you try to call inherit()[0] on your Builder? If that does not work for you, can you share the code that fails?

This crate is probably not easy to use if you do not know how perf works. I was thinking about providing reasonable defaults, but than I am not sure what user are actually looking for.

[0] https://gz.github.io/rust-perfcnt/perfcnt/linux/struct.PerfCounterBuilderLinux.html#method.inherit

[jbreitbart]

@rukai did that solve your problem?