fix: improve Calibre artifact cleanup

[fredchu]

Summary

扩展 clean_calibre_markers() 处理更多 Pandoc/Calibre 转换残留，每条规则附 before/after 样本。

从 PR #2 拆出来的第二项改动（文本清理），按 review 建议修正了激进的全局替换。

改动内容

规则	Before	After
属性块（含 #id）	## Title {#calibre_link-0 .calibre3}	## Title
转义括号包裹	\[Some paragraph.\]	Some paragraph.
空标题	##	(removed)
标题 [*text*] 包裹	## [*Some Title*]	## Some Title

针对 review 意见的修正

1. \[ / \] 不再全局替换 — 改用 (?m)^\\?\\\[ 和 (?m)\\?\\\]$，只匹配行首/行尾，不会破坏 LaTeX display math \[...\]
2. strip('[]* ') 已移除 — 标题清理改用 re.match(r'^\[\s*\*(.+?)\*\s*\]$', text) 精确匹配 [*text*] 模式，不会误删合法字符

Test plan

• 属性块移除：text{.calibre5} → text
• 转义括号：行首行尾的 \[...\] 被移除
• LaTeX 保护：行中的 \[x+1\] 不受影响
• 标题清理：## [*Some Title*] → ## Some Title
• bold 括号：[**Chapter One**] → **Chapter One**

🤖 Generated with Claude Code

[deusyu]

Thanks again for the PR.

After reviewing this against the current main branch, I do not think this is a small patch anymore. Improving Calibre/Pandoc cleanup needs to be handled more carefully under the current regression expectations, so I do not want to merge this directly in its current form.

I am going to close this for now, roll the underlying idea into the roadmap on my side, and handle any follow-up work from the maintainer side.

Thanks again for the contribution.