fix: remove stdsimd feature; add stdarch_x86_avx512

[lazyky]

Closes #61

[jvdd]

Thanks for contributing this @lazyky!

I'm wondering if we could improve this by being backwards compatible with other rustc versions through using cfg_version (or smth similar)..

[codspeed-hq]

CodSpeed Performance Report

Merging #62 will degrade performances by 14.24%

_{Comparing lazyky:main (ae95499) with main (c4a4c3c)}

Summary

⚡ 10 improvements
❌ 11 (👁 11) regressions
✅ 147 untouched benchmarks

Benchmarks breakdown

	Benchmark	main	lazyky:main	Change
⚡	sse_f16_argmax_in	495.5 µs	474.1 µs	+4.51%
⚡	sse_f16_argmin_in	495.6 µs	474 µs	+4.54%
👁	sse_f16_argminmax_in	602.7 µs	624 µs	-3.41%
👁	scalar_f32_argminmax_rn	2.1 ms	2.4 ms	-14.24%
👁	scalar_f64_argminmax_rn	2.2 ms	2.6 ms	-13.26%
⚡	scalar_i16_argmax	1.2 ms	1.1 ms	+11.95%
⚡	scalar_i16_argmin	1.2 ms	1.1 ms	+11.95%
👁	avx2_i32_argmax	338.5 µs	349.2 µs	-3.05%
👁	avx2_i32_argmin	338.5 µs	349.2 µs	-3.04%
👁	impl_i32_argmax	338.7 µs	349.4 µs	-3.06%
👁	impl_i32_argmin	338.8 µs	349.4 µs	-3.04%
⚡	scalar_i8_argmax	1.2 ms	1.1 ms	+11.97%
⚡	scalar_i8_argmin	1.2 ms	1.1 ms	+11.97%
⚡	scalar_u16_argmax	1.2 ms	1.1 ms	+11.95%
⚡	scalar_u16_argmin	1.2 ms	1.1 ms	+11.95%
👁	avx2_u32_argmax	338.6 µs	349.2 µs	-3.03%
👁	avx2_u32_argmin	338.6 µs	349.2 µs	-3.04%
👁	impl_u32_argmax	338.8 µs	349.4 µs	-3.03%
👁	impl_u32_argmin	338.8 µs	349.5 µs	-3.05%
⚡	scalar_u8_argmax	1.2 ms	1.1 ms	+11.97%
...	...	...	...	...

ℹ️ Only the first 20 benchmarks are displayed. Go to the app to view all benchmarks.

[lazyky]

According to rust-lang/stdarch#1486:

• Split the stdsimd feature

[jvdd]

LGTM! If all tests pass, this one will be merged.

(P.S.: sorry for spamming some commits)