Instead of silent data. This shows some difference in the FFmpeg decoded
data, with the data delayed by half a buffer, explaining the additional
data we decode.

The audiodata.bin resource has been generated with:

  gst-launch-1.0 audiomixer name=mix ! \
    audio/x-raw,format=F32LE,rate=22050,channels=2 ! \
      filesink location=dlls/winegstreamer/tests/audiodata.bin \
    audiotestsrc freq=400 volume=0.2 num-buffers=1 \
      samplesperbuffer=22050 ! mix. \
    audiotestsrc freq=600 volume=0.2 num-buffers=1 \
      samplesperbuffer=22050 timestamp-offset=20000000 ! mix.
Signed-off-by: Rémi Bernon <rbernon@codeweavers.com>
Signed-off-by: Nikolay Sivov <nsivov@codeweavers.com>
Signed-off-by: Alexandre Julliard <julliard@winehq.org>