fix vanilla tensor image detection

[pmeier]

We missed that in #5421. Although we often do isinstance(input, torch.Tensor) in the stable implementation, we can no longer do that for the prototypes. Since features._Feature is also a torch.Tensor, everything not explicitly handled before the isinstance(input, torch.Tensor) condition would be treated as image. For example

from torchvision.prototype import features, transforms

sample = dict(label=features.Label(0))
transform = transforms.Resize((32, 32))

transform(sample)

TypeError: Tensor is not a torch image.

[facebook-github-bot]

💊 CI failures summary and remediations

As of commit 260c6fe (more details on the Dr. CI page):

---

💚 💚 Looks good so far! There are no failures yet. 💚 💚

---

This comment was automatically generated by Dr. CI (expand for details).

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.