Add manual `wgpu` API

This will require some changes to how wgpu-native works, namely we'll definitely want to be able to return Bevy's instance rather than requiring users to create a new instance themselves. libprocessing will need to ship with a defined version of wgpu-native that users will be expected to interact with if they want to share wgpu handles between libprocessing and their custom WebGPU code. There's a ton of trickiness here and this will likely be a pretty difficult task.

未指定文件、测试或入口点。首先跟踪 libprocessing 当前如何创建并公开其 Bevy 和 wgpu-native 实例，然后检查预期的 wgpu-native 版本以及句柄共享边界。完成的标准是：在不要求用户创建单独实例的情况下，手动 wgpu API 和实例共享行为已定义并实现。