Harden ingestion workflow and add multi-document readiness

[varun3011]

Summary

Completes docs/task/ingestion-follow-up-hardening-and-validation.md.

This PR hardens the ingestion workflow, improves observability and recovery, adds repeatable load validation, and
extends query readiness for multi-document retrieval.

• Align document status schema with runtime ingestion statuses.
• Add ingestion timing fields and expose them in document APIs.
• Add ingestion health and manual reconciliation endpoints.
• Add stale ingestion reconciliation logic.
• Refresh ingestion run status from worker success/failure paths.
• Add focused tests for retry, reindex, reconciliation, ingestion runs, worker callbacks, and multi-document query
  behavior.
• Add repeatable ingestion load-test tooling and documentation.
• Add load-test artifacts for 10, 25, and 50 valid PDF runs.
• Extend query APIs to support document_ids while keeping existing document_id compatibility.
• Search retrieval chunks across selected documents and log all searched documents.

Additional Frontend Observability Update

• Shows ingestion timing data in the Upload / Ingestion workspace page.
• Adds an Ingestion timing section to Observability.
• Displays average extraction, indexing, total ingestion, and batch wall time.
• Stabilizes document ordering in the Upload page when timestamps match.

Validation

• Real PDF upload/indexing passed.
• 10 valid PDF load test: completed, 10 indexed, 0 failed.
• 25 valid PDF load test: completed, 25 indexed, 0 failed.
• 50 valid PDF load test: completed, 50 indexed, 0 failed.
• Focused backend tests passed: 17 passed.
• Worker regression tests passed: 2 passed.
• Client build passed: npm run build.
• Black checks passed for touched backend/worker files.

Notes

Supabase environments need scripts/schema.supabase.sql applied before using the new ingestion statuses and timing
fields.