Mixed H2/H∞-Policy Learning Synthesis