fix: use Unicode-aware word boundaries for non-ASCII language detection

[dajiaohuang]

Summary

Fixes language detection for non-ASCII scripts by replacing ASCII-only \b word boundaries with Unicode-aware lookarounds using (?<![\p{L}\p{N}]) and (?![\p{L}\p{N}]) with the u (Unicode) flag.

Problem

JavaScript's \b word boundary only works with ASCII letters [a-zA-Z0-9_]. This meant:

• Cyrillic words (Russian) with accented or non-ASCII characters were not correctly matched at word boundaries
• Accented Latin characters (Italian: perché, ciòe; Portuguese: você, código; Spanish: código; etc.) were also not correctly matched

Fix

• Replace all \b boundaries with Unicode-aware lookarounds: (?<![\p{L}\p{N}]) (lookbehind: not preceded by letter/number) and (?![\p{L}\p{N}]) (lookahead: not followed by letter/number)
• Add u flag to all regexes to enable Unicode property escapes
• Update CJK regexes to also use u flag for consistency

Affected languages

• Italian (it)
• Brazilian Portuguese (pt-BR)
• Spanish (es)
• German (de)
• French (fr)
• Russian (ru)
• Indonesian (id)
• Japanese (ja) (flag update only)