Show any-article nearest neighbours and widen the format/kind vocabularies

The article dashboard page gains a "Nearest articles (any)" table: the
ten closest stored embeddings by cosine, regardless of rating or run,
via a brute-force scan of article_embeddings. The deep-assessment format
facet grows from 5 to 14 values (code_repository, documentation_reference,
tool_or_product_page, discussion_thread, paper_or_report,
interview_or_transcript, video_or_podcast, fiction_or_humor, other) and
the triage kind from 10 to 16 (repo, docs, discussion, paper, media,
fiction), so a GitHub repository is no longer forced into analysis_essay.
Both prompt versions bump to 2.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QVPagF6jfDv78CC5Jv2wp4

README.md

@@ -227,6 +227,14 @@ unreferenced or edit it. Custom providers can be removed after no `[llm]` role
 references them. The service needs `/etc/daily-epub` in `ReadWritePaths` for
 these writes.
 
+An article detail page shows the ten closest articles across all stored,
+compatible embeddings, independent of runs or ratings, alongside the rated
+neighbours captured by the latest run. Its triage kind and deep-assessment
+format also distinguish repositories, documentation, product pages,
+discussions, papers, interviews, media, and fiction instead of forcing those
+pages into essay or report labels. Run `features backfill --all` when recent
+articles do not yet have embeddings to compare.
+
 ### Web routes
 
 | Route | Access | Purpose |