Optimal halting policies in Markov population decision chains with constant risk posture