Arm backend: Support channels-last input and output

[pytorchbot]

• Insert transposes for input/output iff the incoming/outgoing data is
  in channels first format.
• For testing using tosa_reference_mode, transpose numpy arrays to and
  from correct data format since numpy doesn't have the concept of
  dim_order.
• Remove checks for channels_first only input.
• Remove check for not changing dim_order before to_tosa_memory_format
  pass since the behaviour of channel last tensors is non-predictable.
• Add dim order testing of example networks and mv2
• Add a section to the documentation about memory formats.

cc @digantdesai @freddan80 @per @zingo @oscarandersson8218

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/14400

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

❌ 1 New Failure

As of commit cb348ae with merge base e0dda90 ():

NEW FAILURE - The following job has failed:

• pull / test-samsung-models-linux / linux-job (gh)
  RuntimeError: Command docker exec -t bb58cb03de5b8685bd4483d9b4009335e906228016a94dafba87ae86c2edd5a6 /exec failed with exit code 1

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[GregoryComer]

Re-running jobs with infra failures. If it looks good, I'll merge shortly.

[GregoryComer]

Samsung failure is unrelated infra. Discussed offline and it's being investigated. Merging.

[digantdesai]

Shouldn't we (somehow) check if the assumptions in the blob about the input tensor dim_order are satisfied? It can silently do wrong compute is my concern.

[digantdesai]

How is it working correctly with NHWC input without mv2.to(channels_last) - it could be because you are assuming every weight is lowered to TOSA and during lowering you might be formatting. But if you can't lower some weight then it may fail, no?