[NisargOza]

WHO landing pages contain only an Overview. This adopts Ashu's WHO implementation from #6 (commit ca7a330fc2e87022e24e6b1b72130b3ec5002408) and ports it onto current main's shared scraper registry, combining the Overview with linked PDF text when conversion succeeds.

• Add the optional Docling pdf extra and root PDF group; record full versus overview content scope and recover from PDF download/conversion failures.
• Reject invalid URLs/limits early, stop safely on later-page listing API failures, and log/skip failed publication pages while counting successful documents.
• Preserve headings, table captions and detected footnotes omitted by Docling's default markdown export. Treat partial conversion as failure.
• Record publisher reuse policy separately from unverified individual-document licensing. Retain attribution for the source code, real PDF excerpts and inherited historical backend measurements.
• Include focused public-entry-point tests and opt-in real-PDF/reference-output checks. Shared scraper modules, NICE and MedlinePlus are unchanged.
• Ignore local corpora, environment variants and handoff notes; document branch names without the codex/ prefix.

Credit: Ashu supplied the original WHO implementation, fixtures and historical five-document backend comparison. The adoption commit includes a co-author trailer. Original #6 remains open. New OCR diagnostics are isolated in #32; CCO is isolated in #30.

Validation: uv run ruff check ., uv run ruff format --check ., AMFV_CHECK_GOLDEN=1 uv run --group pdf pytest -q (88 passed, four third-party deprecation warnings), and uv build --package amfv-datasets. A normal development install without PDF dependencies previously passed 81 tests with seven optional tests skipped. Live checks covered the first listing API page and an Overview-only publication; full-text checks used two six-page real excerpts and mocked HTTP. No full-corpus scrape was performed.