reduce default warnings

Addresses the following warnings:

pytorch_lightning/utilities/parsing.py:198: Attribute 'loss' is an instance of `nn.Module` and is already saved during checkpointing. It is recommended to ignore them using `self.save_hyperparameters(ignore=['loss'])`.
pytorch_lightning/trainer/connectors/data_connector.py:441: The 'val_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.
pytorch_lightning/trainer/connectors/data_connector.py:441: The 'train_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.
pytorch_lightning/trainer/connectors/data_connector.py:441: The 'predict_dataloader' does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` to `num_workers=7` in the `DataLoader` to improve performance.
pytorch_lightning/utilities/data.py:77: Trying to infer the `batch_size` from an ambiguous collection. The batch size we found is 1. To avoid any miscalculations, use `self.log(..., batch_size=batch_size)`.
neuralforecast/tsdataset.py:92: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
  self.temporal = torch.tensor(temporal, dtype=torch.float)
neuralforecast/tsdataset.py:96: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
  self.static = torch.tensor(static, dtype=torch.float)

by doing the following:

Wrapping the save_hyperparameters call in a context manager that ignores warnings because we do want to save the loss passed to the constructor.
Adding pl.disable_possible_user_warnings() in the core module to remove the warnings related to the dataloaders.
Passing the corresponding batch_size to each LightningModule.log call. This is used to weigh the losses across steps in a batch (for the step methods) and across processes (for the epoch methods) in distributed.
Using a copy constructor instead of torch.tensor(some_tensor) in the tsdataset module. I don't follow the suggestion to use detach because these arrays/tensors are always on the CPU, but I could add it.

Also cleans some notebooks.

Nixtla / neuralforecast

reduce default warnings #974