use ffmpeg for sample packing

FFmpeg has a much more optimized algorithm than we do so may as well use it.

app/widget/viewer/viewer.cpp

@@ -434,9 +434,13 @@ void ViewerWidget::ReceivedAudioBufferForPlayback()
     if (watcher->HasResult()) {
       SampleBufferPtr samples = watcher->Get().value<SampleBufferPtr>();
       if (samples && audio_playback_device_) {
-        qint64 t = QDateTime::currentMSecsSinceEpoch();
-        QByteArray pack = samples->toPackedData();
-        qDebug() << "Packing took:" << (QDateTime::currentMSecsSinceEpoch() - t);
+        if (!packed_processor_.IsOpen()) {
+          packed_processor_.Open(samples->audio_params());
+        }
+
+        // Convert to packed data for audio output
+        QByteArray pack = packed_processor_.Convert(samples);
+
         audio_playback_device_->Push(pack);
 
         if (prequeuing_audio_) {
@@ -627,6 +631,7 @@ void ViewerWidget::PauseInternal()
     audio_playback_device_.reset(nullptr);
     qDeleteAll(audio_playback_queue_);
     audio_playback_queue_.clear();
+    packed_processor_.Close();
 
     UpdateTextureFromNode();
   }

app/widget/viewer/viewer.h

@@ -28,6 +28,7 @@
 #include <QTimer>
 #include <QWidget>
 
+#include "audio/packedprocessor.h"
 #include "audiowaveformview.h"
 #include "common/rational.h"
 #include "node/output/viewer/viewer.h"
@@ -257,6 +258,7 @@ private:
   std::unique_ptr<PreviewAudioDevice> audio_playback_device_;
   std::list<RenderTicketWatcher*> audio_playback_queue_;
   rational audio_playback_queue_time_;
+  PackedProcessor packed_processor_;
 
   static QVector<ViewerWidget*> instances_;