System information

• TensorFlow version (you are using): 1.12.0
• Are you willing to contribute it (Yes/No): Not yet decided

Describe the feature and the current behavior/state.
Sometimes we encounter wav files which has JUNK chunk before fmt chunk. CPython's wave module handles this kind of wav files by skipping those unnecessary chunk. But tensorflow's wav_io module expects fmt chunk comes immediately after WAVE tag. As a consequence, those wav files with JUNK chunk, which is playable by media file players, generates Header mismatch InvalidArgument error.

As a conclusion, we need to implement, I may call it as, unnecessary chunk skipping functionality on wav_io.

Will this change the current api? How?
I expect that api won't be affected at all.

Who will benefit with this feature?
Those who are handling wide variety of wav files would be benefit with this feature, as they do not have to do some preprocessing(removing JUNK chunk manually all the time).

Any Other info.
Kaldi toolkit's handling of wav input would be a good reference.