Add LanguageIdentifier::cmp_bytes

[sffc]

Part of #831

I opened #1703 to think about how to scale this to Locale.

[sffc]

Although it wasn't the goal of this PR, if you like what I do here, I could replace other impls like PartialEq<str> and Writeable to use LanguageIdentifierSubtagIterator.

[zbraniecki]

lgtm!

[Manishearth]

lgtm with nit fixed

[sffc]

Although it wasn't the goal of this PR, if you like what I do here, I could replace other impls like PartialEq<str> and Writeable to use LanguageIdentifierSubtagIterator.

I tried it for Writeable and I can't beat the performance from our current Writeable impl, so although it slightly cleans up the code, I'll leave it as-is, at least for LanguageIdentifier. It may be more worthwhile in Locale.

I could simplify PartialEq<str>, but it would change the semantics due to the lenient/strict comparison issue Zibi pointed out. This also means that I can't implement PartialOrd<str> based on cmp_bytes because it has different semantics.

This means I'm adding code without deleting code, which isn't great, but this function has a specific purpose and use case.

[zbraniecki]

optional, n-b:

return self.langid.variants.get(i).map(||);

[sffc]

The main thing to be done inside the map closure would be to update the state variant. I think I prefer using a more procedural code style in this function since it is a nontrivial function and involves mutable state.