Switch to rayon-core for faster build

[silvanshade]

This reduces the build time (with --timings) for blake3 from around 1.7s to 1.3s on a Zen4 7950x.

This helps for situations where one is trying to minimize build time (e.g., for build-dependencies).

There seems to be no real downside since the crate doesn't use anything not included in the core rayon crate.

Below I've included the results from cargo build --timings.

The improvement doesn't really show up for b3sum on my system since other dependencies take long enough that it masks the difference. But it's still probably better to switch them both and maybe in some scenario the difference would surface.

b3sum-rayon-core.pdf
b3sum-rayon.pdf
blake3-rayon-core.pdf
blake3-rayon.pdf

[oconnor663]

Wow yeah, if I do cargo clean && time cargo build on my box, this change brings it down from ~2.0s to ~1.2s. That's a surprisingly big difference. Thanks!