[pipeline] Add pool option to image feature extraction pipeline

[amyeroberts]

What does this PR do?

Adds the flag pool which will return the pooled output, rather than the raw hidden states.

Doesn't work for data2vecvision as the model doesn't add the pooling layer by default. At the moment, model kwargs aren't passed to the model constructor, so it's not simple to add passing add_pooling_layer here.

I chose to raise an error when getting the outputs of the model. Although this means we fail quite late, it avoids any complex inspection needed on the model.

From the list of models in #28944:

Models which work with the pool option:

• beit
• bit
• convnext
• convnextv2
• deit
• dinov2
• dpt
• efficientnet
• focalnet
• levit
• mobilenet_v1
• mobilenet_v2
• mobilevit
• mobilevitv2
• nat
• regnet
• resnet
• swin
• swinv2
• van
• vit
• vit_hybrid
• vivit
• yolos

Models that don't have a pooling layer:

• conditional_detr
• deformable_detr
• deta
• detr
• dinat
• efficientformer
• glpn
• imagegpt
• poolformer
• pvt
• segformer
• siglip_vision_model
• swiftformer
• swin2sr
• table
• timesformer
• timm_backbone
• videomae
• vit_msn
• vitdet
• vit_mae

Not working

• data2vec-vision.

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?

[HuggingFaceDocBuilderDev]

The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.

[ArthurZucker]

LGTM, let's maybe add a pipeline test were we make sure the outputs slices are alright (instead of juste the shape?)

[ArthurZucker]

it more like the pooler_outpus were not returned since we don't check if there is a pooler layer layer or not 😉

[amyeroberts]

@ArthurZucker Updated the error message and added tests for exact outputs in b94d08c

[ArthurZucker]

Nice! Feel free to merge 🤗