Improve the documentation of profiling with "perf maps"

[canova]

The documentation lives here: https://github.com/python/cpython/blob/d6cf05b5d00abbfdf9008d646c14c6ef7c83b61a/Doc/howto/perf_profiling.rst

Before #136461, cpython only had support for Linux perf profiling. But with this PR it now supports both Linux and macOS through samply. I updated the documentation to include this, but it's not super great currently because previously it was written with the assumption that it was only supported by Linux perf. So it would be good to restructure this doc in a better way, and add more information related to macOS profiling using samply.

Also samply supports both perf modes so it would be good to clarify when to use them.

cc @pablogsal @hugovk

[AA-Turner]

Are we likely to get Windows support for 3.15? If so, it would make sense to write in a more general fashion than just POSIX-related operating systems.

A

[canova]

I filed #137022 for the Windows implementation. It looks like it's not terribly difficult to do, but it uses a different approach than jitdump/perf maps.