Verbosity enhancement

[FlorentRamb]

This PR is related to issue #15, with three updates:

1. consider verbose init param in CTGAN as a frequence, i.e. progress results will be printed every verbose epochs
2. print training duration in progress results
3. print device used if verbose is not 0.

[CLAassistant]

All committers have signed the CLA.

[hadjipantelis]

This is really helpful functionality. Is there a chance this same change is done to TVAE too ? TVAE is similarly structured.

[npatki]

Hi @FlorentRamb, thanks for your contribution request. At this time, we are considering using the verbose parameter to show a progress bar using tqdm (similar to the SDV sampling methods). Separately, we may choose to save the loss values per epoch so that they are easily accessible from the model at a future point.

I'll close this PR off since it's not a direction we want to pursue. But do let us know if you're interested in contributing to other features related to verbosity.