Reflect shader stage for bind groups.

[StarArawn]

This PR changes the shader reflection to automatically pick the correct shader stage/s.

[cart]

In general this looks like the right approach. Just left one comment.

[cart]

Testing on my machine now. Then we should be good to go!

[StarArawn]

@cart
After much debugging I have a hack to fix all of the examples. In most of the 3D examples the shader being used is forward.vert/frag which implements the camera uniform the "both" the vertex and fragment stage. The pass node specifies that the resulting uniform must have both a Vertex and a Fragment shader stage. Before when we didn't reflect this wasn't an issue, but now its a bit more of an issue. We need a better mechanism of picking the descriptor perhaps by ignoring shader stage... Perhaps our hash implementation shouldn't account for shader stage at all, but I don't think that's right. I also don't think we should be retrieving the binding descriptor based off of name so I'm a little stuck on how I can "unhack" this.

camera_bind_group_descriptor:
https://github.com/bevyengine/bevy/blob/master/crates/bevy_render/src/render_graph/nodes/pass_node.rs#L73

[multun]

Can you rebase this on master? Format checking was enabled, and I believe the CI doesn't yet check for it at the commit you're at

[StarArawn]

@multun Seems like cargo fmt doesn't work in bevy on stable. I was able to format each file though using vscode/RSA. This is ready for review now.

[cart]

Perhaps our hash implementation shouldn't account for shader stage at all, but I don't think that's right.

Yeah we need the hash to be a 1:1 map between "things wgpu cares about / would require a new wgpu descriptor to work"

I also don't think we should be retrieving the binding descriptor based off of name so I'm a little stuck on how I can "unhack" this.

i think retrieving binding descriptor by index (without checking name) in a cross-shader context is probably even more shakey. what if the shader sets binding 0 to something that isnt the camera? names / conventions give us a set of "rules" we can rely on.

I'm a little stuck on how I can "unhack" this.

Yeah this is a little weird. I think I'm fine with merging this as-is, but we should create a tracking issue to resolve the underlying problem. This probably won't be the last time it comes up.

[cart]

Follow up via #264