…aceOrSolidus (facebook#48828)

Summary:

1. Rename `CSSComponentValueDelimeter` to `CSSDelimeter` bc the names are getting way too long.
2. Make the distinction between `Whitespace` and `OptionalWhitespace`. Note that for property values, and function blocks, the value parser will already remove trailing/leading whitespace, but it's weird that whitespace unlike others was not required to be present
3. Add `CSSDelimeter::CommaOrWhitespaceOrSolidus` for simpler parsing in the common pattern of alpha values, and move CSSColor function parsing to use that

Changelog: [Internal]

Reviewed By: lenaic

Differential Revision: D68461968

Summary:

In the last diff I mixed and matched `<legacy-rgb-syntax>` and `<modern-rgb-syntax>` a bit to keep compatiblity with `normalze-color`.

Spec noncompliant values have only been allowed since facebook#34600 with the main issue being that legacy syntax rgb functions are allowed to use the `/` based alpha syntax, and commas can be mixed with whitespace. This seems like an exceedingly rare real-world scenario (there are currently zero usages of slash syntax in RKJSModules validated by `rgb\([^\)]*/`), so I'm going to instead just follow the spec for more sanity.

Another bit that I missed was that modern RGB functions allow individual components to be `<percentage>` or `<number>` compared to legacy functions which only allow the full function to accept one or the other (`normalize-color` doesn't support `<percentage>` at all), so I fixed that as well.

I started sharing a little bit more of the logic here, to make things more readable when adding more functions.

Changelog: [Internal]

Reviewed By: javache

Differential Revision: D68468275

…#48841)

Summary:

Right now during parsing we can ask for a next component value, with a delimeter, and even if we don't have a component value to consume, we will consume the delimeter.

This is kind of awkward since e.g. trailing comma can be consumed, then we think syntax is valid. Let's try changing this.

Changelog: [Internal]

Reviewed By: lenaic

Differential Revision: D68474739

Summary:

This is a result of inlining isinf previously but we can never be negative infinity because we can never be less than zero.

Changelog: [Internal]

Reviewed By: lenaic

Differential Revision: D68474836

Summary:

Adds support for hsl() and hsla() funtions. This supports more modern syntax options than normalize-color, like number components, optional alpha, and fills in missing support for angle units. The underlying math was lifted pretty much directly from normalize-color though.

An aside, std::remainder for these is not guaranteed to be constexpr, but Clang is still okay with rgb function parsing to be contexpr because we never try to evaluate non-constexpr function?

Changelog: [Internal]

Reviewed By: lenaic

Differential Revision: D68473990

Summary:
The implementation of `hwb()` color functions added in facebook#34600 is pretty flawed.

`hwb()` color functions do not allow comma delimited values. So most of the examples in the unit test here will fail to parse on web. Like `hsl()`, these should also allow numeric non-hue components (instead of just %), and angle values for hue (instead of just numbers), and this is also missing support for alpha values, though these are less dangerous compared to allowing and encouraging incorrect delimiters.

https://www.w3.org/TR/css-color-4/#the-hwb-notation

These were added for web compat, and the examples fail to parse on web, so I'm opting to just remove this incorrect support before implementing this more correctly in the Fabric CSS parser in next diff. I did not attempt to fix the other issues I discovered with the PR implementation in the last couple diffs, around mixing and matching syntax allowed in legacy/modern, along with allowing inconsistent delimiters. 

Changelog:
[General][Breaking] - Remove incorrect hwb() syntax support from normalize-color

Differential Revision: D68591172