Audioflux

* New features:
* Add `audioflux.STFT` algorithm.
* Add `audioflux.PitchFFP` algorithm.
* Add `audioflux.PitchShift` and `audioflux.TimeStretch` algorithms.
* Add method `temporal` to `audioflux.Temporal`.
* Optimize performance.
* Fix bug:
* Fix `nmf` bug.

* Add Cepstrogram algorithm
* Add PitchCEP/PitchHPS/PitchLHS/PitchNCF/PitchPEF/PitchSTFT/PitchYIN algorithm
* Fix bug

* Fix bug:
* Fix the v0.1.5 bug.
* Fix conda bug for macOS.

* New features:
* Optimize macOS/Linux performance.
* Modified API:
* Remove the parameter of `audioflux.mel_spectrogram`: `is_reassign=False`. If you want to use reassign, you can use `audioflux.BFT`.
* Remove the parameter of `audioflux.bark_spectrogram`: `is_reassign=False`. If you want to use reassign, you can use `audioflux.BFT`.
* Remove the parameter of `audioflux.erb_spectrogram`: `is_reassign=False`. If you want to use reassign, you can use `audioflux.BFT`.
* Fix bug:
* Support Ubuntu 18.04 system.

* Python
* New features:
* A variety of audio formats have been supported, such as MP3 and so on
* The macOS ARM platforms have been supported.
* Multi-channel audio has been supported.
* Modified API:
* `audioflux.read` increases support for audio file lists and directory, and add the parameters of monocular conversion and modification sampling rate.
* Modify the `audioflux.Temporal` API.
* Modify `audioflux.CWT` parameter: `num=84`.
* Modify `audioflux.PWT` parameter: `num=84, low_fre=None, high_fre=None`.
* Modify `audioflux.WSST` parameter: `num=84`.
* `audioflux.chirp` use `sicpy` to implement. And modify parameter `phi=None`, remove parameter `linear`, add parameter `method='logarithmic'`.
* Add `audioflux.resample`/`audioflux.convert_mono` method.
* Removed API:
* all debug/enable_debug methods.
* audioflux.Resample.enable_continue
* audioflux.WindowResample.enable_continue
* audioflux.NSGT.get_cell_data
* audioflux.PWT.pwt_det
* audioflux.PWT.enable_det
* audioflux.Temporal.ezr
* audioflux.Temporal.temporal
* Fix bug:
* Memory leakage in `audioflux.Temporal` object.
* `audioflux.WindowResample.__init__` uses C method.
* Fix the known bug.

1. Fix Python3.6/3.7 compatibility issues under Windows
2. Update document