1. Add the ability to work back through queued buffers to find the first buffer that’s older than the given one. This is to enable the replacement of deciphered buffers with buffers that arrive later but are timed earlier. Added functions that relate to this are: get_revised_seqno() to wind back through the queued buffers and clear_buffers_from() to free buffers that are already occupied.

2. Clean up get_audio_buffer_size_and_occupancy() and replace it with get_audio_buffer_occupancy().
3. Add a new general setting: audio_decoded_buffer_desired_length_in_seconds for managing the size of the player buffer queue when the source is buffered audio. Default is 0.75 seconds so that we can step back up to that time if older buffers appear.

common.h

@@ -301,6 +301,8 @@ typedef struct {
   double audio_backend_buffer_interpolation_threshold_in_seconds; // below this, soxr interpolation
                                                                   // will not occur -- it'll be
                                                                   // basic interpolation instead.
+  double audio_decoded_buffer_desired_length;    // the length of the buffer of fully decoded audio
+                                                 // prior to being sent to the output device
   double disable_standby_mode_silence_threshold; // below this, silence will be added to the output
                                                  // buffer
   double disable_standby_mode_silence_scan_interval; // check the threshold this often