[FEAT] Generalisation of conditional WGAN for multivariate timeseries

[pbezz1]

The current implementation of the conditional WGAN supports only univariate datasets. However in many cases where GAN is used to augment timeseries datasets, the data is composed of multiple feature columns. One use case that I am currently working on for an academic research study is the need to generate synthetic financial timeseries of stock market data for multiple assets concurrently and to use the synthetic data in downstream risk management tasks.