[FEATURE] ARM64 + NPU support for Stable Diffusion

[dav3shanahan]

Is your feature request related to a problem? Please describe.

The current AI Dev Gallery lacks native support for Stable Diffusion on ARM64 architecture, specifically for the new generation of Copilot PCs powered by Snapdragon X Elite processors. This limitation prevents developers and users from fully leveraging the powerful NPUs (up to 45 TOPS) in these devices for AI-driven image generation tasks.

Describe the solution you'd like

I request the addition of native ARM64 support for Stable Diffusion in the AI Dev Gallery. This should include:

1. Optimization of Stable Diffusion models for ARM64 architecture
2. Integration with Windows Copilot Runtime for AI acceleration on ARM64 devices
3. Utilization of DirectML optimization tools specifically for ARM64
4. Implementation of ONNX Runtime for efficient model deployment on ARM64 Copilot PCs

Describe alternatives you've considered

Using x86 emulation, which significantly reduces performance and doesn't utilize the NPU
Implementing custom ARM64 builds of Stable Diffusion outside the AI Dev Gallery ecosystem
Utilizing cloud-based Stable Diffusion services, which don't leverage on-device AI capabilities

Additional context

The new Snapdragon X Elite processors offer up to 20x more power for AI workloads compared to previous generations. With 87% of commonly used Windows applications already supporting ARM64 natively, adding Stable Diffusion support would align with Microsoft's vision for AI acceleration at scale and enhance the capabilities of AI-powered Windows devices. This feature would be particularly beneficial for creative professionals and developers working with AI-generated imagery on portable, high-performance Copilot PCs.