[FEA][torchinductor-EVT] tensor construction API that takes in shape + stride directly

[mlazos]

Is your feature request related to a problem? Please describe.
This is a small API change in that is a little annoying to have to check whether strides are row or column major before passing to the tensor constructor at

cutlass/python/cutlass/backend/evt/ir/tensor.py

Line 54 in 5e49724

def __init__(self, tensor=None, element=None, shape=None, layout_tag=None, is_constant=False) -> None:

.
TorchInductor tracks this internally, so it is easier for us to check whether the shape + stride are contiguous and then passing it along.

Describe the solution you'd like
Ideally just take a shape and stride arg rather than having to convert to a shape, layout tag pair and then have the constructor interpret these.

Describe alternatives you've considered
I added a shim layer in inductor that does this, so it's nbd, but just a nice-to-have, perhaps having two versions of the API wouldn't hurt here.

cc @eellison, @henrylhtsang

[hwu36]

@jackkosaian

[jackkosaian]

@apuaaChen, can you take a look at this EVT-related request?