Implement physically based cameras

[Duttenheim]

Currently, the way we calculate luminance and exposure quite arbitrarily. Filament has a very comprehensive section on this topic:

https://google.github.io/filament/Filament.md.html#imagingpipeline/physicallybasedcamera

Basically what it boils down to is that a camera is defined not only by a transform, but also by three optical factors, shutter speed (in seconds), aperture (in f/x) and sensitivity (in ISO), which controls how much light is captured in the final output buffer.