The new "shader stage reflection" currently requires a hack to enable automatic PassNode camera binding

The main issue here is that there are now multiple wgpu "bindings" for cameras because they might be VERTEX | FRAGMENT, just VERTEX, or just FRAGMENT.

This problem will likely pop up in other places too. We need to establish a pattern / method for this use case. The current approach relies on creating bindings ahead of time, before we know the stages a binding will be used in.