[FBCode->GH] Revert "SymIntify roi_align (#7448)"

[NicolasHug]

This reverts commit 8f2e5c9.

The PR seems to be causing issues externally (#7472) and internally (D44552412)

Closes #7472

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/7483

• 📄 Preview Python docs built from this PR

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

❌ 12 Failures, 1 Pending

As of commit d3a2cba:

BROKEN TRUNK - The following jobs failed but were present on the merge base 78c2719:

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

• pytorch/vision / conda-py3_10-cpu (gh)
• pytorch/vision / conda-py3_11-cpu (gh)
• pytorch/vision / conda-py3_8-cpu (gh)
• pytorch/vision / conda-py3_9-cpu (gh)
• pytorch/vision / wheel-py3_10-cpu (gh)
• pytorch/vision / wheel-py3_11-cpu (gh)
• pytorch/vision / wheel-py3_8-cpu (gh)
• pytorch/vision / wheel-py3_9-cpu (gh)
• unittests (3.10, macos-12) / macos-job (gh)
• unittests (3.11, macos-12) / macos-job (gh)
• unittests (3.8, macos-12) / macos-job (gh)
• unittests (3.9, macos-12) / macos-job (gh)

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

[NicolasHug]

I'm gonna mark it as [FBcode->GH] even though it's not strictly correct. This avoids me having to land #7448 in fbcode.

[NicolasHug]

Keeping it as DRAFT to avoid merging for now as there might be a fix for the internal issue

[pmeier]

Stamp