Whenever you're ready, walk me through how you'd approach this.
How would you design and implement an ingestion pipeline using GCP Cloud Functions, Cloud Storage, Pub/Sub, and Cloud Composer? Explain the implementation from file arrival through validation, transformation, orchestration, and curated storage. Provide representative Python and DAG code, and address idempotency, schema validation, retries, dead-letter handling, backfills, security, monitoring, and operational recovery.