Submitting the ability to use SleepQA datasets as PyHealth datasets.

[vvenkatar]

Contributor netids are: vv33, ezralh2 and jl187. The SleepQA paper trains the Facebook DPR framework to train an bi-encoder and a reader model. Our submission includes two classes: one for use to load training data for use by bi-encoder and the second one for use by the reader. We have also included test cases for validating the classes. The code and associated test cases were generated using LLM's. While the assignment talked about extending from BaseDataset object, we chose the SampleDataset object as the sleepqa dataset is more amenable to this object. The BaseDataset object has deep notions of patients, events which doesn't blend well with the training data for SleepQA.