Add support for exporting CLIP submodels in OpenVINO

[msmiatac]

What does this PR do?

• Purpose: Add --submodel {vision,text,full} to the export CLI to allow exporting only the CLIP vision or text encoder. This change stems from the need to simplify CLIP model conversion for the DL Streamer project.
• Motivation: DL Streamer benefits from separate CLIP submodules (vision/text) as standalone IRs. The flag removes downstream splitting complexity and leverages the existing TasksManager registry.
• Implementation: Routes --submodel from the CLI to main_export. For Transformers CLIP under feature-extraction, main_export swaps to model.vision_model or model.text_model so TasksManager resolves the correct clip_vision_model/clip_text_model.
• Documentation: Updates export.mdx with the new flag, usage, and examples for vision-only and text-only exports.
• Tests: Adds CLIPSubmodelExportTest validating that the exported IRs expose the expected inputs (pixel_values for vision, input_ids for text).
• Backward Compatibility: Default behavior remains unchanged (full export if the flag is omitted). No breaking changes.
• Scope: Supports Transformers CLIP for feature-extraction. OpenCLIP has registry coverage and can be similarly wired in a follow-up to map its submodules; not included in this PR.
• Impact: Cleaner exports, reducing integration overhead for DL Streamer and other consumers requiring separate CLIP encoders.

Before submitting

• This PR fixes a typo or improves the docs (you can dismiss the other checks if that's the case).
• Did you make sure to update the documentation with your changes?
• Did you write any new necessary tests?

[IlyasMoutawwakil]

Hi I would suggest using the variant API to control which part of the model gets exported see example of metaclip in optimum-onnx https://github.com/huggingface/optimum-onnx/blob/682988bc2fd972f4594fb66725fa7e7c6c2b628d/optimum/exporters/onnx/model_configs.py#L1280