Removing tensor lists from all APIs

[borisfom]

Back to non-lists API
Good news - it's still script-compliant, too :)

[borisfom]

I am getting those unexpected failures from the backend, do those ring any bells:
E TypeError: tensor_product_uniform_1d_bwd_bwd(): incompatible function arguments. The following argument types are supported:
E 1. tensor_product_uniform_1d_bwd_bwd(num_paths: int, data: ndarray[], math_type: int, inputs: collections.abc.Sequence[ndarray[]], grad
_inputs: collections.abc.Sequence[ndarray[]], outputs: collections.abc.Sequence[ndarray[]], stream: int) -> None
E
E Invoked with types: int, torch.Tensor, int, list, list, list, int
If I roll backend back to the pip version, those go away - so it is some regression introduced since. pip version breaks on *kwargs though (that’s expected).