Jai-sdk

- implementation of report method 211
- added fit method as alias for setup 211
- added append method as alias for add_data 211

Use verbose 2 to get the loss graph, verbose 1 to get only the metrics result.

- Users can now define a batch size for data insertion and prediction when using JAI applications.
- Improved messages during training (reaching an early stopping criteria is one thing, actually detecting it to output it to the user is another).
- Training bar now works in a strictly incremental fashion.
- Pre-processing on `j.fields` deprecated and was removed.

- `fill` method now accepts dataframes with just one column of data (besides the label itself).
- `sanity` method now uses a stratified sampling strategy whenever possible.
- Users can set `patience`, `min_delta` and `max_epochs` in Supervised and SelfSupervised setups using the hyperparams dictionary parameter.

- Users can now download vector representations of their databases. PR 192.

- Users can now define early stopping parameters (`patience` and `min_delta`) when calling `setup`. PR 186
- `fields` method now returns embedded feature names without trailing '_latent' pattern. PR 188
- Interruptions now raise a `KeyboardInterrupt` signal. PR 179

- More flexibility for embeddings on fill and sanity.
- `info` method now returns the size of each database.
- More tests.