stop truncating out-of-bmp identifier escapes in the scanner

[dxbjavid]

Out-of-range identifier escapes get silently truncated to a char

In processUnicodeEscapes a braced unicode escape inside an identifier has its parsed code point cast straight to a char, so only the low 16 bits survive. That means an escape like \u{10041} is quietly accepted as the identifier A, and even values above the U+10FFFF maximum such as \u{110041} slip through the same way, so two identifiers that ought to be distinct (or rejected outright) can end up collapsing onto one. The scanner already rejects \u{10000} here, and the string-literal path range-checks the code point, so this brings identifiers into line by rejecting anything that does not fit in a char rather than aliasing it to a different character.