fix(run): correct SIGINT forwarding

[remimimimimi]

Should fix #4426, but won't solve the problem completely.

We have the following cases:

• In an interactive TTY, the terminal sends SIGINT to the foreground process group.
• In non‑TTY (CI/pipes), the terminal doesn’t deliver SIGINT.
• Windows quirk: Ctrl+C can terminate the parent; we must consume it so only the child handles the console event.
• In interactive runs where the child is in a different process group, the terminal won’t signal the child at all unless the parent forwards.

This PR fixes the first two cases; the third case already had a solution, leaving the fourth. To solve it, we should get PGID of the subprocess (using its PID), but I'm not aware of a way to do this with deno_task_shell. After spending some time looking through its code base, I couldn't figure out any proper way to implement this, so postponed complete solution.

[nichmor]

look good to me at the first glance - will play with it tomorrow.

I wonder, do we need a separate test for this?

[Hofer-Julian]

I wonder, do we need a separate test for this?

Yes, it would be good to make a test out of the nice reproducer in #4426.

If the difference, between non-TTY (CI) and TTY makes this too annoying, I'd be okay with leaving the test for now

[Hofer-Julian]

I can confirm that this fixes the reproducer

[uwu-420]

I just tried this fix on my example from #4532 (comment) and it worked as expected, I only got one SIGINT 👍

[Hofer-Julian]

Thanks Valentin!