Align our PoseDetector API with the TFJS version from Google

[ivelin]

Is your feature request related to a problem? Please describe.
Currently our pose detection API supports multiple models, which is great.

Describe the solution you'd like
It would be nice to align with the TFJS version to make it easier for a broader community of ML researchers to participate in the fall detection project.

Additional context

• Usage Docs
• CreateDetector is the main class where the underlying ML model is selected.
• PoseDetector is the main interface that provides pose estimation given an image input.