Fix compile error for i686-pc-windows-msvc target

[KonaeAkira]

This should (hopefully) fix the sporadic compile failure on CI.

+a lot of trailing whitespace removal done by my editor.

[Lokathor]

This is extremely confusing because the avx512f cfg on the module overall should be preventing this error to begin with. Unless there's actual i686 targets with avx-512? In which case they'd need the imports you adjusted, just from ::core::arch::x86::* instead.

[Lokathor]

Look at the CI... I guess there are i686 targets with the avx-512 flags such that we need to fix up all imports in the avx512 module.

[KonaeAkira]

Looks like avx512f is a supported target feature in i686:

rustc --target i686-pc-windows-msvc --print target-features | grep avx512f

[KonaeAkira]

Yeah... fixing all the doctests is gonna take a while lol

[KonaeAkira]

Turns out doctests have access to the outer scope? So there's no need to explicitly specify core::arch::x86_64 or core::arch::x86 in the doctests. Learned another thing today :)

[Lokathor]

Huh maybe that's new because I remember my doc tests having all sorts of imports when I used to write them a lot.

[KonaeAkira]

Looks like that isn't the case after all. The build_test workflow ran successfully so I thought that was the case. Sorry :/

[KonaeAkira]

I just realized that requiring the user to specify OP from either core::arch::x86 or core::arch::x86_64 isn't really ideal. Isn't avoiding having to manually specify core::arch::* the whole point of the crate?

[Lokathor]

It lets you write, for example, "SSE code" that's portable across both x86 and x86_64 SSE. It's handy enough, even though it's not perfect.

[Lokathor]

Released in 0.9.3