Optional mineru extra, pymupdf backend, and a Pages/Read the Docs site (issues #2, #3, #4)

[mgraffg]

Implements issues #2, #3 and #4 in that order (#3 builds on #2, #4 builds on
#2 and #3), as one run against main at 9956c0b.

#2 — Make mineru an optional extra and add the backend selector

Makes mineru the document2md[mineru] extra instead of a hard dependency,
and adds the BatchConverter(backend=...) / CLI --backend selector
(auto, mineru, pymupdf), with auto resolving to mineru when its CLI
is on PATH and to pymupdf otherwise. __version__ bumped to 0.4.0.

Commit: 470de5e

#3 — Add a pymupdf backend for PDFs with an embedded text layer

Adds document2md/pymupdf_backend.py, reading a born-digital PDF's own
embedded text layer via pymupdf4llm instead of OCR-ing it. Neither backend
silently falls back to the other: a list of page images, or a PDF without
enough of an embedded text layer, under pymupdf raises RuntimeError
naming the mineru extra. Imports are lazy inside functions so the docs job
(installed with --no-deps) can still autodoc the module.

Commit: e612ad7

#4 — Publish a Quarto user site on Pages and developer docs on Read the Docs

Adds a Quarto user site under website/ (install, CLI, Python, backends;
published to ingeotec.github.io/document2md via website.yml on push to
main) and a Sphinx developer site under docs/ (architecture, backends,
development, full API reference; published to
document2md.readthedocs.io). The two sites have two audiences and no
overlapping page.

Commit: ace1751

Verification run on this branch before opening this PR

• python -m pytest tests — 90 passed
• python -m pytest dof2md/tests — 2 passed
• python scripts/check_package_versions.py — document2md 0.3.0 (pypi) → 0.4.0 (local): OK; dof2md: OK: tombstone: final release, not maintained
• python -m sphinx -n -W --keep-going -b html docs/source docs/build/html
  — build succeeded, no warnings
• python -m sphinx -b doctest docs/source docs/build/doctest — 4 tests,
  0 failures
• quarto render website (Quarto 1.9.38) — all 5 pages rendered,
  _site/index.html created

Follow-ups that are not this PR's job

• Read the Docs: nothing in this repository can import the project at
  readthedocs.org — a human has to do that once. Until then
  document2md.readthedocs.io answers 404.
• GitHub Pages: website.yml's render job runs (and passes) on this
  pull request, but the publish job — and therefore the gh-pages branch
  and the live site at ingeotec.github.io/document2md — only exists after
  the first push to main, i.e. after this PR merges. Check the site after
  merging.
• LegalIA's website/pages/document2md.ipynb template still links to the
  old legalia.readthedocs.io API page rather than
  document2md.readthedocs.io; worth a follow-up fix in LegalIA once Read
  the Docs is imported.
• nota2md's ocr extra needs to require document2md[mineru]>=0.4.0
  (this release) before nota2md's next release, or
  pip install nota2md[ocr] breaks for everyone outside the LegalIA repo.

Closes #2
Closes #3
Closes #4

🤖 Generated with Claude Code