Pydub

- Fix bug where you couldn't directly instantiate `pydub.AudioSegment` with `bytes` data in python 3

- pydub will use any ffmpeg/avconv binary that's in the current directory (as reported by `os.getcwd()`) before searching for a system install

- Make it easier to instantiate `pydub.AudioSegment()` directly when creating audio segments from raw audio data (without having to write it to a file first)
- Add `pydub.AudioSegment().get_array_of_samples()` method which returns the samples which make up an audio segment (you should usually prefer this over `pydub.AudioSegment().raw_data`)
- Add `pydub.AudioSegment().raw_data` property which returns the raw audio data for an audio segment as a bytes (python 3) or a bytestring (python 3)
- Allow users to specify frame rate in `pydub.AudioSegment.silent()` constructor

- Add support for RAW audio (basically WAV format, but without wave headers)
- Add a new exception `pydub.exceptions.CouldntDecodeError` to indicate a failure of ffmpeg/avconv to decode a file (as indicated by ffmpeg/avconv exit code)

- Fix a bug in python 3.4 which failed to read wave files with no audio data (should have been audio segments with a duration of 0 ms)

- Fix a bug in `pydub.utils.mediainfo()` that caused inputs containing unescaped characters to raise a runtime error (inputs are not supposed to require escaping)