Fix python custom ops tutorial (#3299)

We tightned invariants in PyTorch 2.7 that exposed a bug with the python
custom ops tutorial. This PR fixes said bug.

Test Plan:
- tested locally

Author: zou3519

advanced_source/python_custom_ops.py

@@ -112,7 +112,10 @@ def crop(pic: torch.Tensor, box: Sequence[int]) -> torch.Tensor:
 def _(pic, box):
     channels = pic.shape[0]
     x0, y0, x1, y1 = box
-    return pic.new_empty(channels, y1 - y0, x1 - x0)
+    result = pic.new_empty(y1 - y0, x1 - x0, channels).permute(2, 0, 1)
+    # The result should have the same metadata (shape/strides/``dtype``/device)
+    # as running the ``crop`` function above.
+    return result
 
 ######################################################################
 # After this, ``crop`` now works without graph breaks: