[Minor BUG] num_workers when calling predict/historical_forecasts of specific models

[pfk-beta]

Hello,

Not sure if it is a bug, or just little inconvenience (darts==0.23.1)

My scenario:

1. calling historic_forecast of TFT model without retrain, and only for few historic elements
2. getting a lot of PossibleUserWarning: The dataloader, predict_dataloader 0, does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` (try 4 which is the number of cpus on this machine) in the `DataLoader` init to improve performance.

After quick debugging:

1. forecasting_model.py(1802)_predict_wrapper() is calling TFTModel's predict function, but doesn't pass num_loader_workers
2. predict_wrapper is called from historic_forecast as almost last statement.

Proposition:

1. Historic_forecast should have extra parameter num_loader_workers - named, or as kwargs...
2. This parameter should be passed to predict-wrapper, and then to predict.
3. Not sure about interfering with other elements of Darts.

Btw. Darts is wonderfull and very usefull software. And big thanks to their developers :)\