Optimal stationary policies in the vector-valued Markov decision process