Revert "Bug 1112438 - Make newCurrentFrameTime more accurate while seeking by checking audio & video timestamp boundary in addition. r=cpearce" r=cpearce

This reverts commit 4b812ca9

The original commit was a work around for poorly encoded videos generated by some FFOS devices which could make the first audio frame non-zero. Trying then to make them appear as zero when seeking to start.

There are better ways to get around this problem.

dom/media/MediaDecoderStateMachine.cpp

@@ -2089,15 +2089,7 @@ MediaDecoderStateMachine::SeekCompleted()
     newCurrentTime = seekTime;
   } else if (HasAudio()) {
     MediaData* audio = AudioQueue().PeekFront();
-    // Though we adjust the newCurrentTime in audio-based, and supplemented
-    // by video. For better UX, should NOT bind the slide position to
-    // the first audio data timestamp directly.
-    // While seeking to a position where there's only either audio or video, or
-    // seeking to a position lies before audio or video, we need to check if
-    // seekTime is bounded in suitable duration. See Bug 1112438.
-    int64_t videoStart = video ? video->mTime : seekTime;
-    int64_t audioStart = audio ? audio->mTime : seekTime;
-    newCurrentTime = std::min(audioStart, videoStart);
+    newCurrentTime = audio ? audio->mTime : seekTime;
   } else {
     newCurrentTime = video ? video->mTime : seekTime;
   }