Text this: Near Optimality of Finite Memory Feedback Policies in Partially Observed Markov Decision Processes