docs(analytics): restore explicit Analytical Judge gate after findings (#357)

[sergstack]

Closes #357.

Summary

Restores an explicit Analytical Judge gate directly after findings in the [Analytics] reasoning/workflow docs — a compact orchestration checkpoint that challenges the analytical conclusion itself before narrative packaging. Docs-only. It reuses the existing controls and adds no second QA framework, taxonomy, method, or intent.

DETERMINISTIC CALCULATION → FINDINGS → ANALYTICAL JUDGE → PASS / REVISE / BLOCKED
→ revise or rerun when required → FINAL FINDINGS → MEMO / REPORT → MEMO QA / JUDGE → ACCEPTANCE

Changes

File	Change
ANALYTICAL_REASONING_STANDARD.md	New §8 Analytical Judge gate — the 7 semantic checks each mapped to an existing control (TASK_PROFILE/intent, METHOD_PLAN+registry, CLAIM_EVIDENCE_REGISTRY lineage, explanation challenge, CONTRADICTING_EVIDENCE+method-disagreement, FINAL_EVIDENCE_SUFFICIENCY+claim ladder, material management synthesis); ANALYTICAL_JUDGE output record; bounded revise/rerun rule; quick runtime collapse. §8→§9 (Runtime collapse), §9→§10 (Stop/escalation) renumbered; full-path block updated
ANALYTICS_WORKFLOW.md	Canonical workflow list + Step 9 make findings → Analytical Judge → pass/revise/blocked → final findings explicit before memo/report
QA_CHECKLIST.md	Post-findings Analytical Judge gate block (orchestration over existing Analysis QA, not a new framework)
ACCEPTANCE_CRITERIA.md	Criterion 12; analytical_judge_status: pass/revise/blocked/not_applicable; blocked-status entry
GOVERNANCE_AND_ANTI_PATTERNS.md	Gate principle; blocker entry; anti-pattern rows — no autonomous retry loop, no second framework/taxonomy, blocked != executed
SMOKE_QA_FOR_ANALYTICS.md	Case forcing a plausible-but-over-strong "root cause" claim to revise (the issue's suggested test); quick-mode collapse case
CHANGELOG.md	Dated entry
ANALYTICS_01/03/05 bundles + provenance audit	Regenerated deterministically

Acceptance criteria (issue #357)

• Workflow explicitly contains findings → Analytical Judge → pass/revise/blocked → final findings
• Judge evaluates conclusions before memo generation
• Seven checks represented semantically without duplicating existing contracts
• claim strength <= final evidence sufficiency remains mandatory
• driver != root cause / correlation != causation remain enforced
• Material contradictory evidence / method disagreement cannot silently pass
• blocked != executed remains enforced
• manual_review_required still governed by the reasoning standard (§1); Judge only surfaces it
• quick mode stays compact — no full Judge record without a material trigger
• Revise/rerun bounded by explicit Judge findings + AES governance
• Smoke QA has an over-strong-conclusion → revise case
• No second QA framework or duplicate taxonomy

Non-goals respected: does not replace Memo Judge or deterministic QA; the 22-method registry, analytical intents, and AES execution governance are untouched; no autonomous agent / self-improvement loop.

Checks

6 canonical scripts + audit_bundle_provenance.py --check + pytest tests/ -q (231 passed).

Rollback

git revert this commit — docs-only; bundles + provenance regenerate from the reverted sources.

🤖 Generated with Claude Code

[github-actions]

Merge gate: protected paths changed. Auto-merge is disabled and owner review is required via CODEOWNERS.