CVE-2026-93990: reject a UTF-16 high surrogate that is not followed by a low
surrogate, by adding LITTLE2_/BIG2_IS_INVALID_CHAR and the matching
little2_/big2_isInvalid4 vtable entries to the UTF-16 tokenizers in
lib/xmltok.c, so a lone high surrogate can no longer consume the following
code unit as a fake low surrogate and hide markup from the parser