The Effective Horizon Explains Deep RL Performance in Stochastic Environments