Wgpu

Changed:

* Update to wgpu-native 0.17.2.1. No changes are needed in downstream code.

In this release the API is aligned with the latest webgpu.idl, and
we updated to wgpu-native (v0.17.0.2).

Added:

* New `wgpu.wgsl_language_features` property, which for now always returns an empty set.
* The `GPUShaderModule.compilation_info` property (and its async version) are replaced with a `get_compilation_info()` method.
* The WebGPU features "bgra8unorm-storage" and "float32-filterable" are now available.

Changed:

* The binary wheels are now based on manylinux 2.28, and the 32bit Linux wheels are no longer built.
* In WGSL: toplevel constants must be defined using `const`, using `let` will now fail.
* In WGSL: it is no longer possible to re-declare an existing variable name.
* Error messages may look a bit different, since wgpu-native now produces nice messages replacing our custom ones.
* Errors produced by a call into a wgpu-native function now produce a Python exception (no more async logging of errors).

Fixed:

* Fixed setting the dpi awareness in the Qt backend, by correctly looking up the Qt version.

Changed:

* Links to readthedocs now point to *stable* instead of *latest*, so that people
reading the docs see these that reflect the latest release.
* Don't enable any features by default (previously WGPUNativeFeature_TEXTURE_ADAPTER_SPECIFIC_FORMAT_FEATURES was enabled).

Fixed:

* Fixed issue related to winid (native widgets) on embedded Qt widgets on Windows (348).
* Fixed our example screenshot tests.

Changed:

* The offscreen `WgpuCanvas.draw()` method now returns a `memoryview` instead of a numpy array.
* The shadertoy util changed internally from using numpy to using a memoryview.

Fixed:

* Fixed that `get_preferred_format()` could crash (in `wgpuSurfaceGetSupportedFormats`) due to an upstream bug in wgpu-native (342)

Added:

* The shadertoy util now supports GLSL, so code from the shadertoy website can be directly copied and run with wgpu (343)