ENNEAGRAM-LLMS-FULL-RELEASE-01: finalize Enneagram llms-full production readback

[fermatmind]

What changed

• Added the final Enneagram llms-full production readback JSON/Markdown report.
• Reconciled ENNEAGRAM-LLMS-FULL-RELEASE-01 manifest/state from pending dependency to readback completed/local-checks-passed evidence.
• Recorded fap-web LLMS Feed Cache Ops run 29198788071 as the cache-only proof for /llms-full.txt.

Why

The llms-full release was previously authorized and readiness was green. This PR closes the train item with auditable production readback evidence only: /llms.txt and /llms-full.txt both contain exactly 116 unique Enneagram canonical URLs, and /llms-full.txt is in complete mode.

Validation

• ruby -e "require 'yaml'; YAML.load_file('docs/codex/pr-train.yaml')"
• python3 -m json.tool docs/codex/pr-train-state.json >/dev/null
• python3 -m json.tool generated/enneagram-llms-full-release/release-readback.json >/dev/null
• Custom readback assertions for /llms.txt=116, /llms-full.txt=116, mode=complete, exact locale/entity distribution, zero blocking safety counters, side effects all 0, and allowed changed-file scope
• git diff --check

Deferred items

• No deployment.
• No CMS body content changes.
• No llms eligibility writes.
• No Search Queue writes or IndexNow submissions.
• /llms-full.txt raw duplicate URL mentions remain advisory for enriched feed content; blocking canonical duplicate count is 0.