Use `wide` to speedup coverage map with stable toolchains

[wtdcode]

I did a benchmark for SIMD implementation of coverage map: https://github.com/wtdcode/libafl_simd_bench

The motivation is that we can use wide to replace std::simd so that we no longer need nightly to gain the speedup. It looks like that wide has almost the same performance as std::simd. I can draft a PR if we agree to introduce wide.

wdyt? @tokatoka @domenukk @rmalmain

[wtdcode]

Attached benchmark from aarch64.

[domenukk]

If it's not 100% the same perf, we can also just use wide for stable and leave simd for nightly

[tokatoka]

it's good

[domenukk]

If it's not 100% the same perf, we can also just use wide for stable and leave simd for nightly

This would also mean we could move to simd in stable if it ever lands, though

[tokatoka]

his repo shows it's almost the same. so no reason to keep portable simd

[domenukk]

Well I think it's good to keep it around in order to switch back some day in the future

[domenukk]

Ppl are still working actively on the feature, and once it lands it'd be better to get rid of the extra dependency I feel. See
rust-lang/rust#86656

[wtdcode]

I will gate different implementations via features so that we still keep that while allowing users to choose.