Include set of "standard" benchmark datasets

[muellermarkus]

Hi everyone,

I really like the idea of the package for quickly trying and evaluating different generative models. However, I think that for developing new generative models, the package is currently somewhat lacking. Compared to the image domain, where well-known benchmark datasets exists (CIFAR, Imagenet, etc.), each paper that proposes a new generative model for tabular data, uses different datasets and often different pre-processing strategies. Instead of just providing different types of data loader, it would therefore be quite important to also provide a certain stock of standard tabular datasets that are already pre-processed. This would also increase the popularity of the package and help establish a common (minimal) benchmark datasets for new generative models for tabular data.