forego memory sample buffer and only write samples that need writing

Rather than writing the entire sequence audio to disk every change, now we
just write the parts we need to write.

app/render/backend/audiorenderbackend.cpp

@@ -28,12 +28,6 @@ void AudioRenderBackend::InvalidateCache(const rational &start_range, const rati
   rational start_range_adj = qMax(rational(0), start_range);
   rational end_range_adj = qMin(viewer_node()->Length(), end_range);
 
-  // Truncate to length if necessary
-  int max_length_in_bytes = params().time_to_bytes(viewer_node()->Length());
-  if (pcm_data_.size() > max_length_in_bytes) {
-    pcm_data_.resize(max_length_in_bytes);
-  }
-
   // Add the range to the list
   cache_queue_.append(TimeRange(start_range_adj, end_range_adj));