Implement String#unicode_normalize and String#unicode_normalized?

[HertzDevil]

Closes #11223.

This is a WIP; the code is in a usable state, but every non-ASCII string performs a full normalization on every call to either method. Ways to optimize these methods will be implemented soon.

The spec file directly downloads the test suite from the Unicode Character Database on each invocation. This is probably better than defining all the ~18k test cases in the spec file itself.

[straight-shoota]

The spec file directly downloads the test suite from the Unicode Character Database on each invocation. This is probably better than defining all the ~18k test cases in the spec file itself.

This is inacceptable. The spec suite must be able to run without any external components.

I think we could either consider incorporating the spec source data into the repository, or make the spec to run manually.

[ysbaddaden]

That would be very nice to have! I just came across the need to remove diacritics from a String, and normalizing to NFD would make it super easy.

I'd just name the methods without the unicode_ prefix (hence #normalize) but that's just a personal preference. I also wouldn't generate a String in #normalized? when the string is maybe normalized, but iterate each char to check if it's normalized or not, but that can be optimized later.

[HertzDevil]

I have no idea how to optimize the slow path for #unicode_normalized? yet. Anyone with a better algorithm in mind could take over.