What About Inputing Policy in Value Function: Policy Representation and Policy-extended Value Function Approximator