Default audio bitrate comes from the last audio track instead of the first #128

When a video has more than one audio track, ff4d reads its default audio bitrate from the last track, not the first.

TwoPass._parse_streams() (twopass.py:340-342) sets audio_stream = stream["index"] for every audio stream it finds, so after the loop it holds the last one. Its docstring and the source_audio_br attribute docs both say it's the first track.

This matters for ShadowPlay and OBS clips, which often have game audio on track 0 and a microphone on track 1. The mic track usually has a lower bitrate, so ff4d picks a lower default audio bitrate than the game audio deserves.

The fix: _parse_streams() already collects every audio stream in self.audio_streams, in order. Read the bitrate from self.audio_streams[0] and remove the returned index and the self.probe["streams"][audio_stream] lookup. Add a test with two audio tracks that have different bitrates.

Fix this together with #127 (the MKV bit_rate crash), since both are in the same few lines.