Update ViTMatte Node

[demoulinv]

This pull request adds an option to limit the inference size in the ViTMatte node, improving flexibility and efficiency when processing images. The main changes introduce a new parameter to enable or disable automatic resizing of input images and trimaps, and adjust the inference pipeline to respect this setting.

Parameter and UI changes:

• Added a new boolean parameter limitInferenceSize to the ViTMatte node, allowing users to enable or disable automatic resizing of input images and trimaps for inference. The inferenceSize parameter is now only enabled if limitInferenceSize is True.

Inference pipeline logic:

• Updated the processChunk method to check the limitInferenceSize parameter and conditionally resize the input image and trimap only when this option is enabled.
• Modified the data preparation for inference: when resizing is disabled, the original image and trimap are used without resizing; when enabled, both are resized to the specified inference size.
• Adjusted the post-processing step to resize the output matte only if resizing was applied during inference, ensuring correct output dimensions.

[Copilot]

Pull request overview

Adds configurable ViTMatte inference resizing while preserving output dimensions.

Changes:

• Adds limitInferenceSize and conditionally enables inferenceSize.
• Supports full-resolution inference when resizing is disabled.
• Conditionally restores matte dimensions after inference.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.