Approximation Algorithms for Partial-Information Based Stochastic Control with Markovian Rewards