Implement work-stealing scheduler

[amezin]

Closes #858

[nicoddemus]

Thanks @amezin!

I will review it in the next few days.

You have made several contributions already, would you like to help maintain the project and become a contributor?

[amezin]

What does "becoming a contributor" mean? GitHub already shows the "Contributor" badge...

[nicoddemus]

Basically write access, being able to review/merge PRs, and making releases.

[nicoddemus]

This looks really good!

Left a few minor comments which would be nice to get in as well.

Did you test this new mode in a large test suite?

[nicoddemus]

Feel free to use type-annotations as well, as we run mypy on the repository. 👍

[amezin]

Did you test this new mode in a large test suite?

https://github.com/ddterm/gnome-shell-extension-ddterm/actions/runs/3843994508/jobs/6546739410

Speed/time seems to be indistinguishable from load - either when the suite is split to multiple jobs, or when I run all tests at once (~4k tests total). On GitHub CI, run to run variance is higher than load vs worksteal difference. When I run the same suite locally, I get maybe 5% speedup compared to load, but nothing really noticeable.

[amezin]

Basically write access, being able to review/merge PRs, and making releases.

Yes, I'm interested. But can't promise I'll be consistently active.

[nicoddemus]

Yes, I'm interested. But can't promise I'll be consistently active.

Sure, no problem, we have no such expectations. 👍

[nicoddemus]

Thanks @amezin!

Will merge this in the next few days to give the chance for others to review too. 👍