Update default trainer to use rich progress bar and history

[BenjaminMidtvedt]

This PR adds QoL to the training.

First, the training history can now always be accessed from the trainer:

model = dl.Regressor(...)

trainer = dl.Trainer(...)
trainer.fit(model, ...)
trainer.history.plot()

Second, we move to using the RichProgressBar as the default. This is because it works much better with notebooks than the default TQDM