A hybrid heuristic and bounding methods for the Max-Mean Dispersion Problem