Fix yolos resizing

[amyeroberts]

What does this PR do?

On main - running the following produces an image outside of the limits set by "longest_edge" in an images config:

from transformers import AutoProcessor
from PIL import Image
import requests

processor = AutoProcessor.from_pretrained("Xenova/yolos-small-300") # or hustvl/yolos-small-300
url = 'https://i.imgur.com/qOp3m0N.png' # very thin image

image = Image.open(requests.get(url, stream=True).raw).convert('RGB')
output = processor(image)
# Result
# main: (3, 89, 1335)
# branch: (3, 80, 1328)
print(output['pixel_values'][0].shape) 

This logic was copied from the DETR image processing logic which comes from the original DETR repo . Note: this means the DETR models' image processors won't respect longest_edge.

Yolos' image processor has been updated to reflect the original model.

Note - this also differers from the output image size that would be obtained from using torchvisions resize. In the above example, it would be (3, 88, 1333)

Fixes #27381

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you read the contributor guideline,
  Pull Request section?
• Was this discussed/approved via a Github issue or the forum? Please add a link
  to it if that's the case.
• Did you make sure to update the documentation with your changes? Here are the
  documentation guidelines, and
  here are tips on formatting docstrings.
• Did you write any new necessary tests?

[ArthurZucker]

Thanks! Let's add a test? 😉