Select chain MDPs of different depths to see how task horizon affects error under function approximation (γ = 0.9 fixed)