Per-event output for `profile`, `offcputime`, `offwaketime`, etc.

I can work on that, just need to check if that's something we want first.

Basically the output of tools which can be used to generate flamegraphs (profile, offcputime, offwaketie, etc.) is aggregated per-event, which is enough to generate flamegraphs, but not to generate heat maps (like the ones generated by [flamescope](https://github.com/Netflix/flamescope)). Would it make sense to also provide a `--perf-format` flag for these tools to have a perf-event output which follows Linux perf output format?

No files or tests are named. Start by locating the implementations of profile, offcputime, and offwaketime, then compare their aggregated output with Linux perf output and the Flamescope expectations. Done means the scope and behavior of a --perf-format option are agreed for the affected tools and the emitted output follows Linux perf format.

Written by the indexing model from the issue text.