Corrected small mistake in Quantile loss function and add type annotation in masked_mae_loss
Adapted Log Loss for regression contexts
These functions enhance our model's flexibility in handling several types of prediction errors specific to time series data.
Detailed description: