feat(wordwrap): support CJK line-breaking rules

[Chronostasys]

Problem

CJK (Chinese, Japanese, Korean) text wrapping is broken. In CJK typography, every character is a valid line-break point — unlike Latin scripts where only spaces and explicit breakpoints allow wrapping. The current implementation treats CJK+Latin sequences without spaces as a single word, causing:

1. Entire mixed-language segments wrap as one unit, wasting half the available line width
2. Text like "manual（手动触发），很可能没跑。" at limit=12 renders as a single long line that overflows, instead of breaking at CJK character boundaries

Before (limit=12)

manual（手动触发），很可能没跑。

The entire string is one "word" (no spaces between CJK chars) → never breaks.

After (limit=12)

manual（手动
触发），很可
能没跑。

Each CJK character is a break point. CJK↔Latin boundaries also break.

Changes

Minimal changes to wordwrap.go Write() method:

• isCJK(r rune) bool: Detects CJK characters by Unicode range (Han, Hiragana, Katakana, Hangul, CJK punctuation, fullwidth forms).
• CJK characters are flushed immediately as individual words, making each one a valid break point — standard CJK typography rule.
• CJK↔non-CJK boundaries trigger a word flush, enabling breaks between scripts (e.g., "这是" | "manual" | "触发").
• Non-CJK behavior is completely unchanged — all existing tests pass.

Test Cases

Added TestWordWrapCJK with 11 cases covering:

• Pure CJK text (each char is a break point)
• CJK mixed with Latin (boundary detection)
• CJK punctuation (fullwidth forms)
• Limit=0 passthrough (no wrap)
• Latin-only (unchanged behavior)

=== RUN   TestWordWrapCJK
--- PASS: TestWordWrapCJK (0.00s)
=== RUN   TestWordWrapCJKNoWrap
--- PASS: TestWordWrapCJKNoWrap (0.00s)
=== RUN   TestWordWrapCJKString
--- PASS: TestWordWrapCJKString (0.00s)

All existing tests also pass (except a pre-existing failure in truncate unrelated to this change).