Create GruD.py

[elhesh1]

Elliot Sherman ehs4

This PR adds a new model to PyHealth called GRU-D, which is designed for handling irregular clinical time-series data. GRU-D uses learned decay to deal with missing values and time gaps. The model is based on the paper by Zhengping Che et al.: https://www.nature.com/articles/s41598-018-24271-9