Machine Learning Feature Store with Automated Data Versioning

Create a PostgreSQL-based feature store for machine learning that supports automated feature versioning, lineage tracking, and point-in-time correct feature retrieval. Implement a system that can handle feature generation, registration, and retrieval with millisecond-level performance. Include mechanisms for feature drift detection, automatic feature deprecation, and integration with model training pipelines.