Reduce .toLowerCase() calls by making compiler aware of string cases.

[Tacodiva]

Add the ability for the compiler to track if a string contains uppercase or lowercase characters, and tries to reduce conversions between uppercase and lowercase strings.

I haven't found any real-world benchmark speedups from this yet, but it does make the compiler produce cleaner code.

[nimeratus]

There are some characters where converting to upper case before comparison is different than converting to lower case, for example the German letter ß (\u00df), which doesn't convert back into itself if you uppercase and then lowercase it

Also, there are some characters that are encoded as a surrogate pair but change when lowercased, which means that joining two strings that are lowercasing-invariant can produce a string that changes when lowercased. Example: 𐐀 (\u{10400} = \uD801\uDC00)