Change NoqaCode to a single string with an offset

[ntBre]

Summary

This PR explores two potential and closely related NoqaCode refactors. The first, in the first two commits, replaces the current NoqaCode(&'static str, &'static str) with NoqaCode(&'static str, usize), where the usize is the index where the prefix and suffix separate. As detailed in this comment, it's not really possible to get a nice, single &'static str, but we can hack around it with a LazyLock to call format! in a static:

ruff/crates/ruff_macros/src/map_codes.rs

Lines 293 to 295 in 7ff37f4

	static CODE: std::sync::LazyLock<String> = std::sync::LazyLock::new(
	|| format!("{}{}", crate::registry::Linter::#linter.common_prefix(), #code));
	NoqaCode(CODE.as_str(), crate::registry::Linter::#linter.common_prefix().len())

In the third commit, I just tried NoqaCode(String, usize), which means NoqaCode can no longer be Copy, but is otherwise a bit more natural.

Test Plan

Codspeed benchmarks on this PR. The &'static str version didn't show any difference, but the String version shows up to a 3% regression. Still, either of these representations should be a bit better if we end up needing to call Rule::from_code since they will avoid having to allocate a new string just to pass it there. A single string should be easier to store on ruff_db::Diagnostic too.

ruff/crates/ruff_linter/src/registry.rs

Lines 19 to 21 in 9cb0243

	impl Rule {
	pub fn from_code(code: &str) -> Result<Self, FromCodeError> {
	let (linter, code) = Linter::parse_code(code).ok_or(FromCodeError::Unknown)?;

[github-actions]

ruff-ecosystem results

Linter (stable)

✅ ecosystem check detected no linter changes.

Linter (preview)

✅ ecosystem check detected no linter changes.

[MichaReiser]

I wonder if this refactor is worth it, considering that it's a slow down. Should we wait until we add secondary_code: String to ruff_db::Diagnostic and then take this up again?