Distributed learning for random vector functional-link networks