Add audio data augmentation process to audio data provider.

[xinghai-sun]

• Add audio data augmentation (injecting noise tracks) to audio data provider.
• Please refer to DS2 paper and DS1 paper for details.
• Please refer to deepspeech.pytorch.
• Refer to the DS2 design doc and update it when necessary.

[chrisxu2016]

add simple design doc https://github.com/chrisxu2016/Paddle/blob/develop/doc/design/speech/README.MD

[chrisxu2016]

construct design doc for augmentation：
https://github.com/chrisxu2016/models/blob/develop/deep_speech_2/augmentation/README.md