Add governed institutional research workflow

[HosungYou]

What changed

• Adds the checkpointed institutional scholarly-database research workflow to the existing LongTable Research surface.
• Implements reproducible metadata ingestion, corpus normalization, lawful PDF Vault handling, screening, and APA 7 manuscript packaging.
• Adds a first-class research-data extraction pipeline: frozen extraction profiles, evidence-backed independent extraction, deterministic conflict detection, adjudication, reliability gates, and extracted-dataset freeze.
• Adds MCP-first decision cards with numbered fallback, receipt-backed hard stops, retry handling, and researcher-facing audit artifacts.
• Documents the separation between demand-corpus document analysis and scholarly supply-corpus review before mismatch analysis.

Why

Researchers need one governed workflow that distinguishes metadata from analytic research data, uses authenticated institutional databases without bypassing access controls, preserves consequential human decisions, and produces manuscript-ready artifacts.

Impact

• longtable research remains the single public workflow surface.
• Analysis and manuscript rendering now require an approved, deterministic extracted-data freeze.
• Extraction profile/hash, reliability, conflicts, adjudications, missingness, and dataset-freeze provenance flow into researcher reports and Methods text.
• Institutional access remains supervised and policy-constrained.

Validation

• npm test
• npm run pack:check
• node scripts/audit-research-artifacts.mjs
• deterministic extraction smoke and full institutional E2E
• Word rendering and visual QA