Pya

* 82, `pyaudio` is now optional: If you plan to use `PyAudioBackend`, install `pya` with `pip install pya[pyaudio]`.
* Fix audio device bug
* 77, Improve code type hint
* 79, Use pyamapping

* Now support Python3.10
* Bugfix 67: When the channels argument of Aserver and Arecorder has not been set it was determined by the default device instead of the actual device.

* Make Aserver a context that can wrap play inside. This context will handle boot() and quit() automatically.
* Move OSX/Linux CI to Github Action from TravisCI.
* Bump numpy dependency, numpy.linspace usage is updated.
* Make Asig.play() arguments explicit.
* Asig.dur property that returns the duration in seconds of the signal

* Add logging helper function.
* Make sure `pya` runs on Python3.8
* Bugfix 43: Handle `None` as a return value of `get_server_info` in `pya.helper.backend.determine_backend` (thanks paulvickers)

* Fix bug of extend mode result in a change of dtype (from the default f32 to f64). Now it is fixed at f32.
* Add convolve() method. A convolution method (default to fft). Now you can apply reverb, do autocorrelation to the signal.
* Update Ugen docstring and remove deprecated variable names
* namespace improvement.
* `sanic` and `notebook` are now optional: If you plan to use `JupyterBackend`, install `pya` with `pip install pya[remote]`

* new Amfcc class for MFCC feature extraction
* new helper.visualization.py and gridplot()
* optimise plot() methods of all classes for better consistency in both style and the API
* new helper functions: padding, is_pow2, next_pow2, round_half_up, rolling_window, signal_to_frame,
magspec, powspec, hz2mel, mel2hz_
* Fixed bug that for multichanels spectrum the rfft is not performed on the x-axis