fix sound forwarding for osx and win32 shadow servers

[Antoine Martin]

Follow up from #391, related to: #899.

When we shadow an OSX or win32 session, we want to forward the sound generated by those applications rather than any capture device (microphone).

It would also be nice to re-use the device selection code from pulseaudio (name match), and maybe introduce better command line arguments for it too.

Some changes already for win32: r12647 should allow us to select the capture device.

On OSX:

• ​How to list available audio i/o devices in Bash: system_profiler SPAudioDataType
• ​AudioObjectGetPropertyData to get a list of input devices
• ​NSSound and Enumerating CoreAudio Output Devices
• ​osxaudiosrc

[Antoine Martin]

Tried using the "wasapi" source and sink plugin since this is the "new and improved" API on MS Windows, but this doesn't seem to work - even with the straightforward gst-launch audiotestsrc ! wasapisink test, so it is still commented out in r15928.
Not much we can do if the latest gstreamer doesn't support it... delaying.

[Antoine Martin]

• wasapisink does work well, we've been shipping it since r15928, it works well enough that it could even be made the default
• wasapisrc: as of r22521 it does look like it's doing something, but nothing comes out...

[Antoine Martin]

Tried wasapisrc again using:

gst-launch-1.0 wasapisrc ! fakesink

And got: Failed to get default device.

So this is not ready.

[migration script]

this ticket has been moved to: ​https://github.com/Xpra-org/xpra/issues/1205