Parameterize autoencoder depth

[jdayton3]

Add layers in a loop for a variable autoencoder (& discriminator?) depth. This would be a good thing to adjust if we have more or fewer samples to combat overfitting, to remove more stuff better, or to speed up computation.