Handle supplementary code points in StringUtils.splitByCharacterType()

[alhudz]

Repro: splitByCharacterType("A" + boldA), where boldA is U+1D400 MATHEMATICAL BOLD CAPITAL A.
Expected: ["A𝐀"], one token, since A and the bold A are both upper-case letters.
Actual: ["A", "𝐀"].
Cause: the shared worker iterates one char at a time and calls Character.getType(char), so each half of a surrogate pair reads as SURROGATE rather than the real category of the code point. Same-type neighbours get split, and in the camelCase path pos - 1 lands inside the pair. splitByCharacterType("5" + boldFive) splits two decimal digits the same way.
Fix: iterate by code point with Character.codePointAt/charCount and classify the whole code point; the camelCase boundary backs up by Character.charCount(Character.codePointBefore(...)). BMP input is unchanged.