[ROCm] Relax tolerance for test_gaussian_blur

[rraminen]

Relaxing the tolerance values to enable the below unit test on AMD gfx90a

test/test_functional_tensor.py::test_gaussian_blur[gaussian_blur-sigma3-ksize2-dt3-large-cuda]

cc @jeffdaily @jithunnair-amd

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/9220

• 📄 Preview Python docs built from this PR

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

✅ You can merge normally! (1 Unrelated Failure)

As of commit 772d781 with merge base e3b5d3a ():

BROKEN TRUNK - The following job failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

• CMake / windows (windows.g5.4xlarge.nvidia.gpu, cuda, 12.6) / windows-job (gh) (trunk failure)
  Process completed with exit code 1.

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