Hamilton-Jacobi Reachability Estimation In Reinforcement Learning