Enable options for AlphaGeometry

This document describes how to enable the four deployment options described in
the architecture: Demo (fallbacks), Local infra (Neo4j/FAISS via Docker), Hosted
services, and Full production (Kubernetes/Ray/CI).

1) Option 1 — Fast demo (no external infra)
 - Create a virtualenv and install requirements-merged.txt
 - Run the demo app locally: python -m backend.run_demo
 - Use the API to create universes and attempt proofs (NeuroSymbolicCore uses fallbacks)

2) Option 2 — Local infra (Neo4j + FAISS)
 - Install Docker Desktop.
 - Start Neo4j using provided compose file:
   docker compose -f docker-compose.neo4j.yml up -d
 - Install FAISS (CPU): pip install faiss-cpu
 - Configure backend/adapters/graph_adapter.Neo4jAdapter connection parameters or use env vars

3) Option 3 — Hosted services
 - Provide connection strings for hosted Neo4j and hosted vector DB in a secure manner (secrets manager)
 - Configure adapters to point to these endpoints (see backend/adapters/*)

4) Option 4 — Full production
 - Build docker images and push to registry
 - Deploy to Kubernetes using manifests in deploy/k8s or Helm using deploy/values.example.yaml
 - Provision Ray or Kubernetes jobs for heavy compute and configure autoscaling

Security notes:
 - Use a secrets manager for tokens and DB credentials. Replace TokenAuth with a proper IAM/OAuth flow in production.