Use singleton for WebGPU instance

[wcandillon]

the idea is to be use that the same instance is always used even in case of hot reload and so on

[Copilot]

Pull Request Overview

This PR refactors the GPU initialization to use a singleton pattern, ensuring that the same GPU instance is re-used even during hot reloads. Key changes include moving the GPU constructor to private, introducing a static getInstance() function, and updating consumers (e.g., RNWebGPUManager) to use the singleton instance.

Reviewed Changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
packages/webgpu/cpp/rnwgpu/api/descriptors/Unions.h	Minor formatting update for readability.
packages/webgpu/cpp/rnwgpu/api/GPU.h	Refactored GPU class to implement singleton pattern.
packages/webgpu/cpp/rnwgpu/api/GPU.cpp	Added static member definitions for the singleton pattern.
packages/webgpu/cpp/rnwgpu/RNWebGPUManager.cpp	Updated usage to fetch the singleton GPU instance.
externals/dawn	Updated subproject commit.

Comments suppressed due to low confidence (1)

packages/webgpu/cpp/rnwgpu/api/GPU.h:76

• The name '_instance' is used both for the static singleton pointer and for the member variable of type wgpu::Instance (line 80), which may lead to confusion. Consider renaming the member variable (e.g., to 'gpuInstance') to improve clarity.

  static std::shared_ptr<GPU> _instance;