Vision: [docs] transforms.ToTensor() doesn't rescale pixels of floating-point images

[asanakoy]

I noticed that ToTensor() only rescales pixels of integer-valued images.
If PIL.mode corresponds to the image with the floating-point representation - no scaling to pixel values will be applied.

I think this should be documented in https://pytorch.org/docs/stable/torchvision/transforms.html?highlight=totensor#torchvision.transforms.ToTensor. Otherwise it's too confusing.

Related to: https://github.com/pytorch/vision/issues/546

[fmassa]

Yes, it should definitely be documented.
As I've mentioned in #546, I think we should do something to mitigate this inconsistency.

Could you send a PR improving the documentation?

[asanakoy]

Yes, sure. I will send a PR.