refactor text classification.

update language model's README

add neural language model.

Refactor the data preprocessor part for DeepSpeech2.

Expand .travis.yml to support docker.

Refactor whole data preprocessor for DS2 (re-design classes, re-organize dir, add augmentaion interfaces etc.).

1. Refactor data preprocessor with new added class AudioSegment, SpeechSegment, TextFeaturizer, AudioFeaturizer, SpeechFeaturizer.
2. Add data augmentation interfaces and class AugmentorBase, AugmentationPipeline, VolumnPerturbAugmentor etc..
3. Seperate normalizer's mean and std computing from training, by adding FeatureNormalizer and a seperate tool compute_mean_std.py.
4. Re-organize directory.

Support variable input batch and SortaGrad.

add tools for model conversion from Caffe to PaddlePaddle

update the text generation demo.

Update librispeech.py for DeepSpeech2.

Add the NCE example.

Add model resuming for DeepSpeech2 trainer.

Summary:
1. Add manifest line check.
2. Avoid re-unpacking if unpacked data already exists.
3. Add full_download (download all 7 sub-datasets of LibriSpeech).