Add Idefics2/3 and SmolVLM Fast image processors + improvements for fast image processors (#38157)

* add working idefics2 fast and improvements for fast nested images processing

* add fast image processors idefics 3 and smolvlm

* cleanup tests

* fic doc idefics2

* PR review and fix issues after merge

* Force providing disable_grouping to group_images_by_shape

* simplify group_images_by_shape

* fix modular

* Fix nits after review

docs/source/en/model_doc/idefics2.md

@@ -162,7 +162,7 @@ To load and run a model using Flash Attention-2, simply change the code snippet
 ```diff
 model = Idefics2ForConditionalGeneration.from_pretrained(
     "HuggingFaceM4/idefics2-8b",
-+    torch_dtype=torch.float16,    
++    torch_dtype=torch.float16,
 +    attn_implementation="flash_attention_2",
 ).to(device)
 ```
@@ -184,7 +184,7 @@ Quantizing a model is as simple as passing a `quantization_config` to the model.
 + )
 model = Idefics2ForConditionalGeneration.from_pretrained(
     "HuggingFaceM4/idefics2-8b",
-+    torch_dtype=torch.float16,    
++    torch_dtype=torch.float16,
 +    quantization_config=quantization_config,
 ).to(device)
 ```
@@ -218,7 +218,10 @@ A list of official Hugging Face and community (indicated by 🌎) resources to h
 [[autodoc]] Idefics2ImageProcessor
     - preprocess
 
+## Idefics2ImageProcessorFast
+[[autodoc]] Idefics2ImageProcessorFast
+    - preprocess
 
 ## Idefics2Processor
 [[autodoc]] Idefics2Processor
-    - __call__
+    - __call__

docs/source/en/model_doc/idefics3.md

@@ -80,6 +80,9 @@ This model was contributed by [amyeroberts](https://huggingface.co/amyeroberts)
 [[autodoc]] Idefics3ImageProcessor
     - preprocess
 
+## Idefics3ImageProcessorFast
+[[autodoc]] Idefics3ImageProcessorFast
+    - preprocess
 
 ## Idefics3Processor
 [[autodoc]] Idefics3Processor

docs/source/en/model_doc/smolvlm.md

@@ -32,7 +32,7 @@ SmolVLM2 is an adaptation of the Idefics3 model with two main differences:
 
 Input images are processed either by upsampling (if resizing is enabled) or at their original resolution. The resizing behavior depends on two parameters: do_resize and size.
 
-Videos should not be upsampled. 
+Videos should not be upsampled.
 
 If `do_resize` is set to `True`, the model resizes images so that the longest edge is 4*512 pixels by default.
 The default resizing behavior can be customized by passing a dictionary to the `size` parameter. For example, `{"longest_edge": 4 * 512}` is the default, but you can change it to a different value if needed.
@@ -192,11 +192,14 @@ print(generated_texts[0])
 [[autodoc]] SmolVLMForConditionalGeneration
     - forward
 
-
 ## SmolVLMImageProcessor
 [[autodoc]] SmolVLMImageProcessor
     - preprocess
 
+## SmolVLMImageProcessorFast
+[[autodoc]] SmolVLMImageProcessorFast
+    - preprocess
+
 ## SmolVLMVideoProcessor
 [[autodoc]] SmolVLMVideoProcessor
     - preprocess