Currently, trimPrefix and trimSuffix in UnicodeString inherit the binary comparison logic from AbstractString without handling Unicode normalization.

This creates an inconsistency where startsWith() returns true for canonically equivalent strings (NFC vs NFD), but trimPrefix() fails to remove the prefix because the bytes differ.

Example:

$nfc = "\xC3\xA9";       // é
$nfd = "\x65\xCC\x81";   // e + ´

$str = new UnicodeString($nfc);

$str->startsWith($nfd); // true
$str->trimPrefix($nfd); // returns "é" instead of "" (current behavior)