Added LSTM depended Policy and USF apprximation to target driven A3C navigation agent

paper - Target Driven Visual Navigation with Hybrid Asynchronous Universal Successor Representations

Trained checkpoints https://drive.google.com/drive/folders/1DNLuxHkwhHLVprL9R_b8mmhDaIpAgOJe?usp=sharing

please refer the follow up paper- VUSFA-Variational-Universal-Successor-Features-Approximator and code