Implement the dynamics model

[radiachkik]

The dynamics model is used to representate the dynamics of the environment. It is used for the tree search, because it returns the next hidden state and immeditate reward for a given state-action pair (s, a). That means that in f.e. atari games, it tries to learn the dynamics of the game and in chess f.e. it learns the rules and also the opponent's style of play.