Add Prometheus metrics endpoint

[StephanMeijer]

Context

Find currently has basic health checks via Dockerflow middleware (__heartbeat__, __lbheartbeat__) and an internal self-test framework (core/selftests.py) that collects timing data for database, cache, and OpenSearch connectivity. However, there is no Prometheus-compatible metrics endpoint for scraping operational metrics.

For production observability, we need a /metrics endpoint that exposes standard application metrics in Prometheus format.

Requirements

• Add prometheus-client library to dependencies in src/backend/pyproject.toml
• Create a /metrics endpoint that exposes:
  • Request counts and latencies by endpoint (documents/index/, documents/search/, documents/delete/)
  • HTTP response status code distribution
  • OpenSearch query latencies
  • Document indexing throughput (count, bulk size)
  • Search query performance (query time, result counts)
• Consider adding middleware or decorators to instrument existing views in core/views.py
• Ensure the metrics endpoint is excluded from authentication requirements
• Document the available metrics and scrape configuration

Technical notes

• The existing self-test framework in core/selftests_builtin.py already collects duration_ms for each health check - these could be exposed as gauges
• Gunicorn configuration is in docker/files/usr/local/etc/gunicorn/find.py if process-level metrics are needed
• Consider using prometheus_client.multiprocess mode for multi-worker Gunicorn deployments