Quantum Tensor Networks for Variational Reinforcement Learning