A high-order Discontinuous Galerkin Method with mesh refinement for optimal control