Fix speedscope time interval

[gjoseph92]

This fixes two things:

1. With multiple threads/processes, each may be a different length, so the end_value needs to be set independently. (Also before, the end_value was mistakenly the number of threads, not the number of samples in a thread.)
2. weights are (apparently) not unitless, and need to be given in seconds.

Closes #430

Didn't have time to add tests, but I could if you'd like.

FWIW I was manually testing with:

import threading
import time

threads = [threading.Thread(target=time.sleep, args=(i / 2,)) for i in range(4)]
for t in threads:
    t.start()

for t in threads:
    t.join()

[benfred]

looks good - thanks! I tested myself before/after and the output makes a bunch more sense now

(btw - keep up the good work on dask! I use dask for work and love it =).

[gjoseph92]

Keep up the good work on py-spy! It's been invaluable lately. We know you're busy but we'd love to nerd-snipe you on dask/distributed#4825 and dask/distributed#4987 someday :)

[gjoseph92]

Hey @benfred whenever you have the time, it would help us a ton if you could cut a new py-spy release with this change. I'm profiling multithreaded things for dask, and having the nonsensical times is a little annoying.

[benfred]

@gjoseph92 I've tagged a new release - which will be pushed out to pypi/cargo etc with by this CI run https://github.com/benfred/py-spy/actions/runs/1216029122 when it finishes

[gjoseph92]

Thank you!!