Scalable Learning of Safety Guarantees for Autonomous Systems using Hamilton-Jacobi Reachability.