Basically, mpg123 assumptions about audio is broken. It assumes that
audio ioctl that change mode are instantaneous, which they are not, most
often, so that with the old model, start-up was 80 ioctl, which amounts to
~10 seconds with wss.
Added lazyness to basic audio.c so that it won't try to determine whether
we can play @ 8000Hz mono when 44100Hz stereo is what we want, and it
works...