Initialize Agent abstraction

[lherman-cs]

MediaEngine <---MediaFrame----> AgentActor <--> Network

AgentActor is platform independent and handles networking, WebRTC, and high-level statistics.
MediaEngine is platform dependent and handles native media capture, and codec (which may be hardware-accelerated or software).
MediaFrame is an abstract data structure to contain each media frame:

1. For video, each frame is 1 MediaFrame, e.g. 1frame per 30fps.
2. For audio, each frame may be 20ms, 40ms, 60ms, or even 100ms.

MediaEngine will contain adapters for .ivf/.h264 & .ogg to MediaFrame for testing or pre-recorded use cases. Typical pipelines will encode each raw frame and pass the encoded frame.

[lherman-cs]

will separate MediaEngine work into a separate issue