Story: Document deduplication recipe

As the doc graph grows, duplicate documents appear — two files that cover the same ground under slightly different names. Leaving both alive creates confusion about which is authoritative and breaks the principle that every id-link resolves to exactly one canonical source.

This story delivers a recipe that generalises the deduplication procedure: given two documents, identify the candidate for survival (richer, more complete, better linked) and the candidate for deletion; merge any unique content from the deletion candidate into the survivor; delete the deletion candidate; update all id-links that pointed to the deleted UUID.

The worked example that drives the recipe is a pair of near-identical sprint health review runbooks discovered in doc/llm/runbooks/.

• [ ] doc/recipes/documentation/how_do_i_deduplicate_two_documents.org exists, #+version: 2, scaffolded via generate_v2_doc.sh.
• [ ] Recipe covers all four steps: nominate survivor/deletion, merge unique content, delete the deletion candidate, update all id-links.
• [ ] Recipe is wired into the Documentation recipes index.
• [ ] The duplicate run_a_sprint_health_review runbook is deleted; its unique chart-regeneration content is merged into run_sprint_health_review.
• [ ] doc/llm/runbooks/runbooks.org catalogue row updated to point to the survivor UUID (30FE3C0F).

• Automated duplicate detection — finding duplicates is manual; this recipe only governs what to do once two have been identified.
• Deduplication of non-org files (source code, scripts).