Generate from webgpu.yml

[eliemichel]

The official repository of webgpu.h now provides a machine-processable version of webgpu.h, namely webgpu.yml.

Ultimately, we should generate this WebGPU wrapper directly from webgpu.yml, but this means dropping support for older versions that do not have a webgpu.yml coming with them. And more importantly, this is not possible yet because implem-specific version of webgpu.h that are used everywhere do not come with their own webgpu.yml for now.

One possibility would be to turn the current dirty scraper into a script that extracts a minimal webgpu.yml from any webgpu.h, then have the main generator script use either this reconstructed webgpu.yml or the official one.

[zokrezyl]

Hi, was about to ask the question, why the code is not generated from the yaml file, but understand that the yaml file is something new.

Even if the yaml file would not be used a bit more elegant solution would be to use clang index python lib or C++ clang index lib.

It is also strange that the headers repo does not mention how the yaml file is derived from the idl (webgpu-native/webgpu-headers#484)