In #71 we added the ability to add post-processing effects to the rendered image.

renderer.post_processing_steps.append(NoisyPostProcessingStep(renderer.device))

We should give this API some more thought. Maybe alternatives work better.

One idea could be to just use our existing tools, and create a new renderer that takes the result of the first renderer and renders it using geometry that maps it to the full viewport. But that will result in a lot of boilerplate?

Another idea: these post processing classes, like NoisyPostProcessingStep look a lot like materials, in that they define the shaders. Maybe they can even be specific kind of materials?

In all of this we should be aware that these post-processing steps are specific to the wgpu renderer backend.