[FEA] Tensor manipulation methods for `cute.Tensor`

[nandor]

Which component requires the feature?

CuTe DSL

Feature Request

Is your feature request related to a problem? Please describe.

It would be useful for cute.Tensor to support permute, transpose, unsqueeze, expand. Calling them in host-side launchers instead of torch would reduce CPU-side overheads.

Describe the solution you'd like

Expand the cute.Tensor class to include the methods suggested.

Describe alternatives you've considered

Defining helper methods

Additional context

N/A

[fengxie]

Yes, I think we can create some examples to demonstrate how to manipulate layout in host jit function so we can compile them into nop.

Here is one example:
https://github.com/NVIDIA/cutlass/blob/main/examples/python/CuTeDSL/cute/blackwell/kernel/dense_gemm/dense_gemm_persistent.py#L1399