Orchestrate the end-to-end review run across L0-L3, including chunking, cost estimation, progress, telemetry, dryrun, and auto-approval.
Central coordinator with adaptive chunking, optional preview execution, structured reporting, and metrics hooks around each stage.
Load config/session, chunk the diff, run L0-L3 per chunk, aggregate results, evaluate auto-approve, and emit the final report.
Calls every core layer and consumes shared utilities plus concurrency helpers; provides the main API used by CLI/invocations.