demux_raw: output smaller audio packets

Currently, audio packets are always filtered as a whole. Since demux_raw
output a 1 second long packet, this could lead to large delays when
applying softvol volume. It could be fixed by splitting the frames the
decoder outputs before filtering them (like the old filter code used
to), but since this didn't cause any other problems yet, I'm going with
the simpler fix.

Fixes #1558.

demux/demux_raw.c

@@ -148,7 +148,7 @@ static int demux_rawaudio_open(demuxer_t *demuxer, enum demux_check check)
     *p = (struct priv) {
         .frame_size = samplesize * sh_audio->channels.num,
         .frame_rate = sh_audio->samplerate,
-        .read_frames = sh_audio->samplerate,
+        .read_frames = sh_audio->samplerate / 8,
     };
 
     return 0;