Set # of threads to use performant cores (#2352)

Summary:
Pull Request resolved: https://github.com/pytorch/executorch/pull/2352

When using all cores, slower ones are dragging the performance down by blocking
large cores. Perhaps when we have uarch specific implementation, we may not
need this, but this tool is useful in general until we have better API

//unrelated failures
bypass-github-export-checks

ghstack-source-id: 218918368
exported-using-ghexport

Reviewed By: digantdesai, kirklandsign

Differential Revision: D54766071

fbshipit-source-id: f49afc2afeb16f73ebc10b43c97d1f4a97e3d191