Converge face detection metadata between shape detection API and VideoFrame

[youennf]

We would ideally converge these two APIs to use similar metadata.
At least from an editor point of view, we should reuse enums and types as much as possible between the two.