Implement continuous voice activity detection #1118

Problem Description

Right now, the VAD module supports a static inference on given audio chunk. However, it lacks native support for continuous run i.e. streaming. As a user, having to implement it on JS/React side is not so pretty of a task.

What should be done

Implement streaming mode within the native, C++ side, similar to the one that Whisper models have.

Benefits to React Native ExecuTorch

Continuous VAD would open the door to implementing advanced speech mode - where user does not need to perform any action to trigger speech recognition. This, combined with speech to text module, would have countless number of applications - from voice-controlled interfaces, to personal AI assistants.