macros/class: Propagate cfg annotations to COM class methods

[MarijnS95]

We have a very peculiar workaround for DXC (pending upstream fix solution microsoft/DirectXShaderCompiler#3793) that requires us to conditionally provide function implementations for extra vtable "spacers": https://github.com/Traverse-Research/hassle-rs/blob/f5a090c70bbcf66f3bafd3d549716a414e873838/src/wrapper.rs#L130-L139. The spacers are conditionally defined in: https://github.com/Traverse-Research/hassle-rs/blob/f5a090c70bbcf66f3bafd3d549716a414e873838/src/unknown.rs.

These need to be forwarded otherwise the initialization of these vtable members will happen unconditionally even when they don't exist: https://github.com/Traverse-Research/hassle-rs/actions/runs/1777800733

Fixes #166

[MarijnS95]

Alternatively we could have that pub(crate) type IDxcUnknownShim = IUnknown; but it requires propagating the attribute on the impl instead.

[jelmansouri]

Hi! Is there any update on this PR, our game engine has a dependency in our shader building infra which is Linux only to the DirectXShaderCompiler and we use hassle-rs. Legacy non Microsoft Com library is unmaintained and still causing us issues. Having DirectXShaderCompiler + Rust working properly is important for the professional game development community which uses HLSL most of the time, especially that Rust is taking more and more it's place in the tooling side games!

[MarijnS95]

@rylev can we get a review, merge and release for #235 and this PR (#234)? That would make us able to start using com-rs in production on hassle-rs 🎉

[MarijnS95]

@rylev not sure if you still intend to at least merge outstanding PRs and make another release (and at also push the missing tags to this repo), but I don't really care about following up on this anymore as the upstream fix has finally made its way in, so I can use published com 0.6 now rather than this branch on a fork.