Supermodularity and Affine Policies in Dynamic Robust Optimization