Decentralized and Uncoordinated Learning of Stable Matchings: A Game-Theoretic Approach