Add actor network training system

[generic-github-user]

Optimize the parameters of the actor model to maximize the predicted reward value generated by the critic.