Follow-up from "fmp4: Return a running time in get_next_time()"

The following discussion from !1127 (merged) should be addressed:

• @slomo started a discussion: (+1 comment)

  I think this is actually wrong. chunk_start_pts already is a running time! peek_buffer() is converting the PTS to running time already and then queueing up GOPs with that. All the PTS/DTS in this code after peek_buffer() is always in running time unless it's suffixed with _position (e.g. earliest_pts_position).