Allow v2 Normalize to reuse tensor statistics

[medisean]

Summary

Fixes #9576.

• Allow transforms.v2.Normalize to receive tensor mean/std values directly.
• Preserve tensor objects instead of converting them to Python lists, so torch.as_tensor in the kernel can reuse already device-resident statistics without repeated host-to-device copies.
• Keep sequence inputs backward compatible and add regression coverage.

Testing

• python3 -m py_compile torchvision/transforms/v2/_misc.py test/test_transforms_v2.py
• git diff --check
• The focused PyTorch test is deferred to CI because PyTorch is not installed in this checkout.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/9621

• 📄 Preview Python docs built from this PR

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

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