Always plot data in 2D

[agitter]

Currently, the software only visualizes a dataset if it is two dimensional. We could apply PCA or UMAP for dimension reduction so that we can always visualize the data and decision boundary. We would need to decide what to do with discrete features, for instance, visualizing the one hot-encoded version.