refactor: Modify audio codec API to use AudioBuffer

[yezhan10]

What's included in this PR:

1. All audio encoders consume AudioBuffer, return AudioFrame pointer.
2. All audio decoders consume AudioFrame, return AudioBuffer shared pointer obj. Since it's possible that a audio frame has multiple decoded PCM data. So split the original Decode() function to Decode() and GetDecodedAudioFrame().
3. Added unit tests for AudioPipe, OpusEncoder.

Timestamp improvement will be included in next PR.

[murillo128]

In order to be able to progress this PR, can we split it in two PRs?

The first one about method changes to use the AudioBuffer and the second one to fix the timestamp calculation inside the AudioPipe?

[yezhan10]

In order to be able to progress this PR, can we split it in two PRs?

The first one about method changes to use the AudioBuffer and the second one to fix the timestamp calculation inside the AudioPipe?

@murillo128 @bcostdolby I have updated the implementation to use a queue of audio buffers in audio pipe, and also added timestamp calculation for audio buffer based on decoder pts. if we all like this approach better, i will clean up the code a bit in my next commit. This version still has Encode(const AudioBuffer::const_shared& audioBuffer, BYTE* out, int outLen), I will update this to return an encoded frame in next PR most likely. Also, maybe we can use audiobuffer pool as an improvement in the next PR as well to avoid frequent audio buffer creation.

[murillo128]

looking better, still some changes I would like to make

[murillo128]

lgtm