Linear Program Approximations for Factored Continuous-State Markov Decision Processes