Maximum Entropy RL (Provably) Solves Some Robust RL Problems