RegExp: interaction of splitting large Unicode chars with quantifiers and charsets

[lth]

The "Update Unicode" proposal says that a 16-bit ECMAScript implementation must split a literal like \u{1AFFE} into two code points (a surrogate pair). Necessarily this will also be true in regular expressions.

Then quantifiers applied to such a character probably need to take that into account: \u{1AFFE}* is perhaps (?:\u{1AFFE})*, for example, the two characters acting as an atomic sequence during matching.

A harder problem is character sets, [...\u{1AFFE}...], where the splitting probably does not have the desired effect. It is possible to imagine the meaning of this to be (?:(?:\u{1AFFE})|[...]).

We need to investigate this more fully and explicate the meaning of splitting characters in regular expressions.

[lth]

Also see #213.