Codex iterative repair loop cookbook for Qdrant embedding search — modernize embedding model and query path

AI Impact Summary

This addition provides a Codex iterative repair loop cookbook with a compact Qdrant embedding search fixture, enabling rapid validation of repair workflows. It demonstrates a one-pass cleanup to modernize the local Qdrant query path and clarifies fixture framing, surfacing issues like a legacy embedding model (text-embedding-ada-002) and a deprecated qdrant.search usage. The included repair_story and issue_layers give engineers a concrete path for incremental improvements, accelerating reliability testing for embedding-based search migrations.