[Torch] Support SmoothL1 Loss function in torch dialect

[zahidwx]

SmoothL1 Loss function

https://docs.pytorch.org/docs/stable/generated/torch.nn.SmoothL1Loss.html

Smooth L1 loss function can be implemented either by:

1. decomposing it to torch primitive ops
2. using huber loss function

[zahidwx]

I would like to contribute to this issue.