fix: 修复尾随和声拆分及注音、翻译对齐边界

[imsyy]

修复尾随和声左括号位于词中间时正文被划入和声的问题，保留括号前的文本及其时间跨度，并支持左括号位于第一个词内、同词包含主唱与和声、日文注音保护等情况。对缺少词内时间信息的片段沿用原始时间跨度。

同时完善 #5、#7、#8 的边界行为：数字串的注音覆盖整个数字串；QRC/KRC 外部 kana 在康熙部首归一化前对齐；LRC 过滤翻译占位符后仍保留罗马音的字段位置。

新增 12 个回归用例，包含 #4 的真实 QRC 样例。验证通过：229 个测试、pnpm lint、pnpm typecheck、pnpm build。

Closes #4

[Copilot]

Copilot review overview

🔵 Needs a closer look

LRC placeholder handling can still misplace romanization when multiple placeholders share a timestamp.

Review effort: Lite
Findings: None

What changed in this PR

Fixes trailing harmony splitting and kana/translation alignment edge cases across QRC, KRC, and LRC parsing.

Changes:

• Improves in-word harmony splitting and timing preservation.
• Corrects numeric kana and external kana alignment.
• Adds regression tests for reported edge cases.

File	Summary
tests/​parsing-regressions.spec.ts	Adds regression coverage.
src/​utils/​kana.ts	Fixes kana timing across numeric strings.
src/​utils/​bg.ts	Improves in-word trailing harmony splitting.
src/​parse/​lrc.ts	Handles translation and romanization slot alignment.
src/​parse/​index.ts	Defers Kangxi normalization during kana alignment.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.