Add training support for SigLIP (#31495)

* Add siglip loss function

* Update docs

* Enable training tests
[experimental] enable GC training tests as it has worked for my own data

* Remove test_training* overrides to enable training tests
[run_slow] siglip

* Skip training tests for Siglip text model and ImageClassificationModel
[run_slow] siglip

* Skip GC training tests for SiglipForImageClassification

* Explicitly skip training tests for SiglipVisionModel
Add skip reason for training tests for SiglipTextModel

* Remove copied from to fix CI

docs/source/en/model_doc/siglip.md

@@ -27,7 +27,7 @@ The abstract from the paper is the following:
 ## Usage tips
 
 - Usage of SigLIP is similar to [CLIP](clip). The main difference is the training loss, which does not require a global view of all the pairwise similarities of images and texts within a batch. One needs to apply the sigmoid activation function to the logits, rather than the softmax.
-- Training is not yet supported. If you want to fine-tune SigLIP or train from scratch, refer to the loss function from [OpenCLIP](https://github.com/mlfoundations/open_clip/blob/73ad04ae7fb93ede1c02dc9040a828634cb1edf1/src/open_clip/loss.py#L307), which leverages various `torch.distributed` utilities.
+- Training is supported but does not use `torch.distributed` utilities which may limit the scalability of batch size. However, DDP and FDSP works on single-node multi-gpu setup.
 - When using the standalone [`SiglipTokenizer`] or [`SiglipProcessor`], make sure to pass `padding="max_length"` as that's how the model was trained.
 - To get the same results as the pipeline, a prompt template of "This is a photo of {label}." should be used.