Add LSVI with UCB for linear mdps

See #1

README.md

@@ -190,6 +190,7 @@
 * **`UCRL2`** [Near-optimal Regret Bounds for Reinforcement Learning](http://www.jmlr.org/papers/volume11/jaksch10a/jaksch10a.pdf), Jaksch T. (2010).
 * **`UCBVI`** [Minimax Regret Bounds for Reinforcement Learning](http://proceedings.mlr.press/v70/azar17a.html), Azar M., Osband I., Munos R. (2017).
 * **`Q-Learning-UCB`** [Is Q-Learning Provably Efficient?](https://papers.nips.cc/paper/7735-is-q-learning-provably-efficient), Jin C., Allen-Zhu Z., Bubeck S., Jordan M. (2018).
+* **`LSVI-UCB`** [Provably Efficient Reinforcement Learning with Linear Function Approximation](https://arxiv.org/abs/1907.05388), Jin C., Yang Z., Wang Z., Jordan M. (2019).
 
 ### With a Generative Model