Cleanup prototype kernels for degenerate inputs

[pmeier]

Follow-up to #6542. Resolves

• fix padding for degenerate segmentation masks #6542 (comment)
• fix padding for degenerate segmentation masks #6542 (review)
  Although cloning here is a O(1) operation since there is no data to copy, it also does nothing useful. Whether or not we clone or not, the Tensor.data_ptr() stays the same and we only create another Python object.
• fix padding for degenerate segmentation masks #6542 (review)

The only thing left are the xfailed reference tests introduced in #6542. As explained there, these are not bugs in the kernel, but rather in the test setup specifically the expected output computation. @vfdev-5 Could you look into these?

[pmeier]

Test failure is unrelated:

____________ test_eager_vs_scripted[gaussian_blur_image_tensor-50] _____________
Traceback (most recent call last):
  File "/home/runner/work/vision/vision/test/test_prototype_transforms_functional.py", line 751, in test_eager_vs_scripted
    torch.testing.assert_close(eager, scripted)
  File "/opt/hostedtoolcache/Python/3.7.13/x64/lib/python3.7/site-packages/torch/testing/_comparison.py", line 1359, in assert_close
    msg=msg,
  File "/opt/hostedtoolcache/Python/3.7.13/x64/lib/python3.7/site-packages/torch/testing/_comparison.py", line 1093, in assert_equal
    raise error_metas[0].to_error(msg)
AssertionError: Tensor-likes are not equal!

Mismatched elements: 1 / 1024 (0.1%)
Greatest absolute difference: 1 at index (2, 0, 13, 9)
Greatest relative difference: 0.011627906933426857 at index (2, 0, 13, 9)

[datumbox]

LGTM, thanks @pmeier. Nice clean ups.

The failing tests seems unrelated to me, I think this is known to be flaky. Might be worth fixing.