Add support for passing pytorch tensors directly to kernel calls

[NaderAlAwar]

This checks if torch is installed and uses torch.is_tensor() to check if the passed argument is a pytorch tensor.

[gliga]

If torch is not available, will it print a user friendly error message?

[NaderAlAwar]

Not at the moment but this is not restricted to PyTorch. We need to have better error messages for invalid arguments passed to kernels.