feat: run the same formatter in parallel for many files

[infinisil]

From the website I'd have thought this would be done already, but after running it over Nixpkgs it turns out it's not.

Parallel formatting seems to originally have been implemented by @basile-henry in https://github.com/numtide/treefmt/pull/17, which subsequently didn't make it in @zimbatm's https://github.com/numtide/treefmt/pull/68. While https://github.com/numtide/treefmt/pull/71 attempted to re-implement it, it only runs each formatter in parallel, but not individual files for the same formatter.

This PR fixes that, making each chunk of 1024 files run in parallel (so no extra parallelism if there's less than that number of files).

I tested this with

[formatter.nixfmt-rfc-style]
command = "nixfmt"
includes = [ "*.nix" ]

with https://github.com/NixOS/nixfmt on https://github.com/nixos/nixpkgs/ and can confirm that it sped it up considerably.

---

This work is sponsored by Antithesis :sparkles:

[zimbatm]

Nice. Do you know if pariter still respects the number of threads when invoked recursively? This change could have implications for machines that have access to less resources.

[infinisil]

No more threads are spawned than available cores: https://github.com/rayon-rs/rayon/blob/main/FAQ.md#how-many-threads-will-rayon-spawn