Fix: os.nanosleep for Windows

[kassane]

Close: #8488

Test:
https://zig.godbolt.org/z/4xYbKso9q

[matu3ba]

Quick hint that we should use nanosleep for both Windows + posix, because Linux offers no better guarantee than Windows: https://frenchfries.net/paul/dfly/nanosleep.html

See https://forum.powerbasic.com/forum/user-to-user-discussions/source-code/56824-ntdelayexecution-sleep-sleepex-with-higher-resolution-absolute-delay and https://www.winehq.org/pipermail/wine-devel/2020-August/171640.html and https://stackoverflow.com/questions/54582249/64bit-precision-sleep-function and https://stackoverflow.com/questions/7827062/is-there-a-windows-equivalent-of-nanosleep for context.

Accuracy is roughly ~15ms depending on system clock, so I think calling it nanosleep is misleading on non-realtime systems (afaiu, only realtime systems provide something close to nanoseconds sleep accuracy.) and we could consider a different name.

Please correct me, if and where I might understand it wrong.

Any opinion @kassane ?

[kassane]

Wow! Regarding #2571 the nanosleep function C API syntax sugar across operating systems.

--- edit

Finishing reading the references, there may be a need to rethink the issue of nanosleep.
Second point, it is noticed in the vision of @andrewrk that the conventional sleep of zig is an ideal abstraction between the systems. I agree that the abstraction idea is very good.
However, it does not solve the problem, as some applications need the signal emitted by the system to carry out the timing interruption. I had a bad experience with libVLC binding that ignores zig time.sleep and works with os.nanosleep.

[kassane]

This didn't fix the issue with the libvlc sleep time.
kassane/libvlc-zig#10