docs: use PostgreSQL with a persistent volume in the kind guide

[franciscojavierarceo]

Summary

Follow-up to #182. The kind guide's warning told readers to use PostgreSQL and a persistent volume for real deployments while the example itself used pod-local SQLite. This makes the documented path match the advice:

• Add a Deploy PostgreSQL section: postgres:17-alpine Deployment with a PersistentVolumeClaim on kind's default standard StorageClass, and a Secret carrying both the password and the full DATABASE_URL.
• Point the Agentic API Deployment's DATABASE_URL at the Secret instead of the SQLite literal; note that schema migrations run automatically on startup.
• Add a Verify persistence section: delete the Agentic API pod and continue the same conversation via previous_response_id afterwards.
• Cover the new resources in cleanup and keep SQLite documented as the no-persistence smoke-test fallback.

Test Plan

Executed end-to-end on a Linux arm64 host (kind + Docker) against a live vLLM server:

• PostgreSQL Deployment reached ready; Agentic API rolled out with the Secret-provided DATABASE_URL and /ready returned 200 (migrations ran on startup).
• Created a response containing a codeword, continued it via previous_response_id (recalled correctly), deleted the Agentic API pod, and continued again after the rollout: the conversation survived, and the codeword was recalled.
• Deleted the PostgreSQL pod as well; after its rollout the same continuation still worked, confirming the PVC preserves the data directory.
• Embedded Kubernetes manifests validated with a YAML parser; pre-commit hooks pass on the changed file.

AI assistance (Claude) was used for this change; every line was reviewed by the submitter.

🤖 Generated with Claude Code

[maralbahari]

@franciscojavierarceo can you fix DCO?

[franciscojavierarceo]

@franciscojavierarceo can you fix DCO?

Done!