We previously had an error when a mono capture device was used with
a stereo codec. This is prevented by avoiding any remixing in
AudioProcessing. Instead, capture side downmixing is now done before
resampling. Upmixing can now be handled properly by AudioCoding,
since the AudioProcessing error condition has been removed.
On the render side, downmixing now occurs before resampling. Ideally
this would be handled still earlier in the chain. Similarly, downmixing
for the AudioProcessing reference data occurs before resampling. This
code has been refactored into RemixAndResample, with a comprehensive
unittest added in output_mixer_unittest.cc.
BUG=issue624
TEST=manually through voe_cmd_test, by using mono and stereo capture
and render devices with mono and stereo codecs. voice_engine_unittest,
voe_auto_test.
Review URL: https://webrtc-codereview.appspot.com/676004
git-svn-id: http://webrtc.googlecode.com/svn/trunk@2448 4adac7df-926f-26a2-2b94-8c16560cd09d
