fix(ansi): treat wide characters as breakable in Wrap and Wordwrap

[tzengyuxio]

Summary

• Treat wide characters (CJK ideographs, fullwidth chars) as individually breakable in Wrap() and Wordwrap() functions
• Per Unicode line breaking rules (UAX feat: join strings with a conjunction in a handful of languages #14), breaks are allowed between any two wide characters — they should not be accumulated into a single "word"

Problem

Previously, consecutive wide characters were accumulated into the word buffer and treated as a single long "word". This caused suboptimal wrapping for CJK text:

// width=36, input: "...這個 vault 不只是記事本，而是一個可瀏覽..."

// Before: "vault" alone on a line, CJK text hard-wrapped separately
模板、屬性與 Bases 結合後，這個
vault
不只是記事本，而是一個可瀏覽、可過濾
、可重組的個人知識資料庫。

// After: "vault" shares line with CJK, natural character-by-character wrapping
模板、屬性與 Bases 結合後，這個
vault 不只是記事本，而是一個可瀏覽、
可過濾、可重組的個人知識資料庫。

Fix

When a wide character (width > 1) is encountered, flush the current word buffer and write the character directly to output. This allows line breaks between any two wide characters, producing more natural wrapping for CJK text while preserving word boundaries for narrow (Latin) characters.

This fix benefits all consumers including glamour (via lipgloss.Wrap) and glow.

Related: charmbracelet/glamour#185

Test plan

• All existing tests pass
• Updated 2 test expectations to reflect improved CJK wrapping behavior
• Verified with mixed CJK/Latin text at various widths