Backend: OpenSearch integration (Vector + KV)

[jongalloway]

This issue tracks adding OpenSearch as a backend supporting both vector search and key-value storage where applicable.

Goals

• Implement OpenSearch-backed vector search for NLWebNet abstractions.
• Evaluate/reuse KV (document storage) where it simplifies demo scenarios.
• Provide configuration and environment setup for local (docker) and hosted OpenSearch.

Tasks

• Create NLWebNet.Backend.OpenSearch implementing IDataBackend and vector operations.
• Configuration schema and binding (endpoint, credentials, index mapping, dims, metric).
• Service registration extension and sample config entries.
• Integration tests using docker-compose or Testcontainers.
• Documentation: setup, performance notes, TLS/cert guidance, troubleshooting.
• Wire into demo sample with feature flag.

Acceptance Criteria

• OpenSearch backend can be enabled via config and used alongside other backends.
• Vector insert/search APIs work and pass integration tests.
• Sample + docs allow local run with Docker.

Link

• Parent: 🗄️ Advanced Backend Support #40 Advanced Backend Support