Extend vertical partitioned demonstration

[TTitcombe]

Description

Extend the MVP (partitioning MNIST into images and labels) to work on arbitrary vertically partitioned datasets

Why?

The dataset/dataloader/data partitioning/splitNN architecture/PSI functions are coded assuming the provided data is MNIST and the partitioning function split images and labels. Fortunately, the real world has more data than just MNIST. In this epic we will generalise the code to work with many datasets and partitions

Breakdown

IN PROGRESS

Who else?

May require work in PySyft and PSI

Additional Context

~Should be completed after #2 and #3~