Include more docstrings

[almarklein]

We have docstrings for most methods, but it would be nice to also have docs for the emums/flags/structs and their fields.

The docstrings for the methods are written by hand, the codegen only updates signatures in these files. Would probably be good to read through them and make them more complete ...

For flags/enums/structs we could perhaps load these from upstream definitions. Or just write them by hand in a separate document. Or a mix of both.

Possible sources to load docstrings from:

• It would have been nice if the IDL file contained docs, but it does not.
• https://github.com/webgpu-native/webgpu-headers/blob/main/webgpu.yml (also see Use webgpu.yml for parsing #460). This defines the webgpu.h but most docstrings can be applied to the user-facing API too. However, most docs are currently missing.
• Just scrape https://gpuweb.github.io/gpuweb
• The text for WebGPU spec seems to be defided in https://github.com/gpuweb/gpuweb/blob/main/spec/index.bs

[Vipitis]

Something that might be helpful: cross-refs within the documentation as well as links to the webgpu spec or wgpu docs page

[Vipitis]

just noticed: some type information in docstrings for encoder methods is outdated and misleading, example:

wgpu-py/wgpu/_classes.py

Lines 2007 to 2019 in e36ad2a

	# IDL: undefined copyTextureToBuffer( GPUTexelCopyTextureInfo source, GPUTexelCopyBufferInfo destination, GPUExtent3D copySize);
	def copy_texture_to_buffer(
	self,
	source: structs.TexelCopyTextureInfoStruct | None = None,
	destination: structs.TexelCopyBufferInfoStruct | None = None,
	copy_size: tuple[int, int, int] | structs.Extent3DStruct | None = None,
	) -> None:
	"""Copy the contents of a texture (view) to a buffer.
	Arguments:
	source (GPUTexture): A dict with fields: texture, mip_level, origin.
	destination (GPUBuffer): A dict with fields: buffer, offset, bytes_per_row, rows_per_image.
	copy_size (int): The number of bytes to copy.

[Vipitis]

so we don't forget: there is now a lot more docstrings in the webgpu.h and wgpu.h headers, so we might be able to use most of that (perhaps minutes the init marco stuff).