[enhancement]: Feature Request: Support for SeedVR2 GGUF models in the Upscaling tabIs

### Is there an existing issue for this?

- [x] I have searched the existing issues

### Contact Details

_No response_

### What should this feature add?

Currently, the Upscaling tab in InvokeAI explicitly states that it only supports SD 1.5 and SDXL model architectures. This limits workflows when working with newer, powerful generation models like Flux.1 or 2, Z-Image, or Qwen.

I would like to request support for the **SeedVR2** architecture, specifically optimized GGUF versions such as **seedvr2_ema_7b-Q4_K_M.gguf** and **seedvr2_ema_7b_sharp-Q4_K_M.gguf**. This integration will enable high-quality upscaling, detail restoration, and sharpening directly in InvokeAI for images created using non-SD architectures.

Using external tools or manual workflows to upscaling images created with Flux/Qwen compromises the integrity and usability of InvokeAI.

Why **SeedVR2 GGUF** is a great fit for InvokeAI: Low video memory footprint: Q4 versions are approximately 4.5 GB, making them readily available to virtually all users and consumer hardware. Proven quality: I have extensively tested **SeedVR2 (both Q4 and Q8 quantization)** in both hobbyist and commercial projects. Delivers exceptional results in sharpening, detail enhancement, and restoration of lost detail. Hardware performance: tested on a laptop with an RTX 4070 Ti (8GB VRAM) and a desktop with an RTX 5070 Ti (16GB VRAM) - the model performs flawlessly and efficiently on both configurations.

### Alternatives

Alternatively, you can use free programs like Upscayl or run the image through ComfyUI using the same SeedVR2 model, but integrating this feature into InvokeAI would speed up image processing.

### Additional Content

_No response_