The Multiarmed Bandit Problem Under the Mean-Variance Setting