Docs: fix SigLIP2 usage examples (#43641)
* Docs: fix SigLIP2 usage examples
Use torch_dtype instead of dtype in pipeline/from_pretrained examples and keep processor/model IDs consistent in the quantization snippet.
Ref: #39692
* Use dtype in SigLIP2 pipeline example
* Use dtype in SigLIP2 AutoModel examples
---------
Co-authored-by: Fahad Alghanim <fkalghan@email.sc.edu>
