fix: slice search sections on the same line breaks markdown-it counts

[Abhishek-B-R]

Summary

• project_sections splits the body with a regex on \r\n, \r and \n instead of str.splitlines().
• regression test in tests/test_wiki_sections.py.

Why

heading boundaries come from markdown-it's token.map, which only counts \n, \r\n and \r. splitlines() also breaks on form feeds, U+2028, U+2029, U+0085 and a few control chars, so one of those in a page shifts every section after it. the section body then starts with its own ## heading line and search reports the wrong heading for a match.

Verification

uv run pytest            # 565 passed
uv run ruff check .      # clean
git diff --check

Docs and wiki

• README/docs updated if user-facing behavior changed.
• .almanac/ wiki updated if an implementation decision, workflow, invariant, or gotcha changed.
• Not applicable.

the new test fails on main and passes here.

Notes for reviewers

CRLF and lone \r bodies split the same as before. ruff format --check flags a trailing blank line in sections.py that is already on main; left it alone to keep the diff small.

---

^{Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.}