This should fix the following puplic issues:

  * https://github.com/mackron/miniaudio/issues/106
  * https://github.com/mackron/miniaudio/issues/187

Public issue https://github.com/mackron/miniaudio/issues/126

Public issue https://github.com/mackron/miniaudio/issues/126

Public issue https://github.com/mackron/miniaudio/issues/126

Public issue https://github.com/mackron/miniaudio/issues/126

Public issue https://github.com/mackron/miniaudio/issues/126

Public issue https://github.com/mackron/miniaudio/issues/213

  * The main loop is now handled properly by pa_threaded_mainloop.
  * Rather than waiting for operations to complete inline, the main
    loop is now iterated on a separate thread.
  * Data is now written and read to and from the relevant stream via
    callbacks rather than a hacky loop.
  * Code overall has been simplified.
  * Includes a rant about bad API design in PulseAudio.

This should hopefully address these public issues:

  * https://github.com/mackron/miniaudio/issues/106
  * https://github.com/mackron/miniaudio/issues/187

The PulseAudio mainloop and context objects have been moved out of the
device and into the context.

Public issue https://github.com/mackron/miniaudio/issues/201

Public issue https://github.com/mackron/miniaudio/issues/209

This uses the "enabled" notion rather than "available" as I think it's
a bit clearer in terms of enabled at compile time rather than available
at run time.

Public issue https://github.com/mackron/miniaudio/issues/211

The following APIs have been added:

  * ma_is_backend_available()
  * ma_get_avaialable_backends()

Public issue https://github.com/mackron/miniaudio/issues/211

Public issue https://github.com/mackron/miniaudio/issues/201

This is a simplified and somewhat more intuitive way of handling fades.
With these APIs, fades are applied immediately, whereas with the old
fading APIs the fades were scheduled. The old APIs still exist, but may
be removed.

The ma_sound_set_fade_in_frames/milliseconds() API sets the fade which
will be applied immediately. If the starting volume is negative, the
fade will start from the current fade volume. This will also overwrite
any existing fade that is already happening.

The ma_sound_get_current_fade_volume() API allows you to retrieve the
current volume of the fade.

The fade is layered on top of the normal volume parameter. I.e. they
are configured and applied to the signal interdependently.

Currently, all fades are applied linearly.