Pydub

- Add the ability to add a cover image to MP3 exports via the `cover` keyword argument to `pydub.AudioSegment().export()`
- Add `pydub.AudioSegment().get_dc_offset()` and `pydub.AudioSegment().remove_dc_offset()` which allow detection and removal of DC offset in audio files.
- Minor fixes for windows users

- Make `pydub.AudioSegment()._spawn()` accept array.array instances containing audio samples

- Make `pydub.AudioSegment()` objects playable inline in ipython notebooks.
- Add scipy powered high pass, low pass, and band pass filters, which can be high order filters (they take `order` as a keyword argument). They are used for `pydub.AudioSegment().high_pass_filter()`, `pydub.AudioSegment().low_pass_filter()`, `pydub.AudioSegment().band_pass_filter()` when the `pydub.scipy_effects` module is imported.
- Fix minor bug in `pydub.silence.detect_silence()`

- Update `pydub.AudioSegment()._spawn()` method to allow user subclassing of `pydub.AudioSegment`
- Add a workaround for incorrect duration reporting of some mp3 files on macOS

- Add support for radd (basically, allow `sum()` to operate on an iterable of `pydub.AudioSegment()` objects)
- Fix bug in 24-bit wav support (understatement. It didn't work right at all the first time)

- Add support for python 3.5 (overstatement. We just added python 3.5 to CI and it worked 😄)
- Add native support for 24-bit wav files (ffmpeg/avconv not required)