Distributionally Robust Policy and Lyapunov-Certificate Learning