Add seven-part production spine

[Tashima-Tarsh]

Summary

• add official/open-source parser readiness registry for priority sources: CAG, finance, GST, NCRB, CERT-In, Gazette, LGD, WRIS, and NDMA
• add claim-level provenance checks so dashboard claims cannot publish without parser-backed records
• add seven-capability production spine report and protected /api/v1/production/readiness endpoint
• extend database schema for durable evidence events, source ingestion runs, and claim provenance
• include the seven-part OpenAI/open-source operating mode in the premium re-architecture doc

OpenAI / open-source boundary

• Open-source operation remains deterministic with no model provider
• OpenAI is optional through DISHA_MODEL_PROVIDER=openai and OPENAI_API_KEY
• OpenAI may summarize evidence but may not invent source rows, statistics, government facts, or dashboard values

Verification

• npm.cmd --prefix web run type-check
• npm.cmd --prefix web run test -- production-spine.test.ts architecture-control-plane.test.ts model-intelligence.test.ts
• npm.cmd run verify