fix(serializer): render relative hyperlinks with portable separators

[eren-YU]

Fixes docling-project/docling#4220.

What changed

TextItem.hyperlink is typed AnyUrl | Path. A Path hyperlink read out of a document is a native path, so on Windows str() yields backslash separators and every serializer that stringified it emitted a non-portable destination:

exporter	before
Markdown	[next page](sub\next.html)
HTML	<a href="sub\next.html">
LaTeX	\href{sub\textbackslash{}next.html}
DocLang	<href uri="sub\next.html"/>

and save_as_json wrote "sub\\next.html", so a document converted on Windows loaded on POSIX as a single path component containing a backslash.

A shared hyperlink_uri() helper now renders a Path destination with as_posix(), and the four serializers that write a hyperlink destination use it.

Why not the existing image-path encoder

MarkdownPictureSerializer._escape_uri_path() is the obvious-looking reuse, but it percent-encodes URL delimiters because it assumes a filesystem path: sub/next.html#section becomes sub/next.html%23section and sub/next.html?x=1#section becomes sub/next.html%3Fx=1%23section. That changes the link target rather than just its separator, so the fragment/query delimiters would stop working.

as_posix() only swaps the separator and leaves everything else — ?, #, % escapes and spaces — untouched.

How I checked

Exported a document with Path and AnyUrl hyperlinks on Windows:

input	before	after
Path("sub/next.html")	sub\next.html	sub/next.html
Path("sub/next.html#section")	sub\next.html#section	sub/next.html#section
Path("sub/next.html?x=1#section")	sub\next.html?x=1#section	sub/next.html?x=1#section
Path("a%20b/x.html")	a%20b\x.html	a%20b/x.html
Path("dir with space/x.html")	dir with space\x.html	dir with space/x.html
AnyUrl("https://example.com/a/b?q=1#f")	unchanged	unchanged

11 tests added to tests/test_serialization.py covering the helper, Markdown and HTML export, the fragment/query cases from the issue discussion, and a JSON round trip. All 11 fail before the change and pass after.

uv run pytest tests/test_serialization.py tests/test_plain_text_serialization.py -q

→ 119 passed. No new failures elsewhere in the suite. ruff check and ruff format --check are clean.

[github-actions]

✅ DCO Check Passed

Thanks @eren-YU, all your commits are properly signed off. 🎉

[mergify]

Merge Protections

🔴 1 of 2 protections blocking · waiting on 👀 reviews

	Protection	Waiting on
🔴	Require two reviewer for test updates	👀 reviews
🟢	Enforce conventional commit	—

🔴 Require two reviewer for test updates

Waiting for

• #approved-reviews-by >= 2

This rule is failing.

When test data is updated, we require two reviewers

• #approved-reviews-by >= 2

---

Show 1 satisfied protection

🟢 Enforce conventional commit

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|docs|style|refactor|perf|test|build|ci|chore|revert)(?:\(.+\))?(!)?: