feat: add MetaGPT Valkey cookbook

[atao2004]

Summary

Add a runnable MetaGPT Valkey integration cookbook covering setup, vector-store RAG, and production deployment.

Changes Made

• Add getting-started, vector-store RAG, and production guides.
  • Explains the proposed MetaGPT integration from feat: add Valkey as RAG vector store backend FoundationAgents/MetaGPT#2063.
  • Documents JSON vector storage, HNSW/FLAT indexing, KNN search, TLS, ACLs, tuning, monitoring, and lifecycle management.
• Add a standalone ValkeyVectorStore sample using valkey-glide-sync.
  • Keeps the cookbook runnable while MetaGPT PR #2063 remains open and unreleased.
  • Uses lazy connection creation, configurable timeouts, TLS/authentication, atomic batches, safe SCAN-based cleanup, dimension validation, and safe vector query parameters.
  • Catches GLIDE RequestError in check_connection() and includes a regression test.
• Add executable sample entry point, dependency list, environment example, Docker Compose setup, and live integration tests.
• Add cookbook Python CI workflow.

Testing

• 10 live integration tests passed against Valkey Bundle.
• End-to-end sample passed.
• Markdown lint passed.
• Python syntax validation and git diff --check passed.
• Focused pytest rerun was unavailable in the final shell because llama_index was not installed; the regression test is included for CI.

Upstream Status

The cookbook tracks FoundationAgents/MetaGPT#2063, which is still open and unmerged.

Notes

The sample intentionally does not depend on an unreleased MetaGPT package. The sequential SCAN + JSON.GET deletion path is retained because it supports matching arbitrary source-document IDs and is documented in the guide.

[jbrinkman]

Closing this draft PR as part of a cookbook program cleanup.