Implement custom q value parser according to spec

[jannschu]

No description provided.

[Copilot]

Pull Request Overview

Implements a new QValue type to parse and compare HTTP quality values according to the RFC specification, replacing the previous f32-based logic.

• Replaces f32 quality fields with QValue and updates comparisons in best_match and sorting routines.
• Adds a QValue struct with FromStr, Ord, and utility methods plus comprehensive unit tests.
• Refactors parse_q_value to return QValue with unwrap_or_default() for missing parameters.

Comments suppressed due to low confidence (3)

src/lib.rs:156

• Add a unit test for Accept::best_match or equivalent that verifies media types with q=0 are correctly excluded by this branch.

                    if quality.is_zero() {

src/lib.rs:215

• Introduce a test that confirms media types are sorted by specificity and then by QValue, to ensure the new sort_by_key behavior matches expectations.

        media_types.sort_by_key(|x| {

src/lib.rs:224

• [nitpick] The method name parse_q_value now returns a QValue type instead of f32; consider renaming it to parse_qvalue or updating its doc comment to reflect the new return type more clearly.

    fn parse_q_value(media_type: &MediaTypeBuf) -> QValue {

[Copilot]

[nitpick] Consider extracting the nested parse_fractional function out of the FromStr impl into a private helper at module scope for clarity and easier unit testing.

Suggested change

	fn parse_fractional(digits: &[u8]) -> Result<u16, HeaderError> {
	digits
	.iter()
	.try_fold(0u16, |acc, &c| {
	if c.is_ascii_digit() {
	Some(acc * 10 + (c - b'0') as u16)
	} else {
	None
	}
	})
	.map(|num| match digits.len() {
	1 => num * 100,
	2 => num * 10,
	_ => num,
	})
	.ok_or_else(HeaderError::invalid)
	}

[jannschu]

Could also inline the function since it is only called once

[maxcountryman]

I'm not putting a lot of stock in what the AI says, I'm mostly just curious if it's at all intelligent.

[maxcountryman]

Thanks for this. We can cut a new release inclusive of this and the last patch. I'll bump the minor version since this is technically breaking and we're not yet 1.0.

[jannschu]

Sure, thanks!