Revert of Always post buffering state updates since they may change s…

…ink state. (patchset chromium#1 id:1 of https://codereview.chromium.org/1161463006/)

Reason for revert:
Speculative revert for Blink layout tests crashes on Nexus 4:
http://build.chromium.org/p/chromium.webkit/builders/WebKit%20Android%20%28Nexus4%29
(SEGV on CrGpuMain thread is observed)

Sorry if my guess is not correct; there wasn't
a clear signal to identify the culprit.

Original issue's description:
> Always post buffering state updates since they may change sink state.
>
> The lock must never be held when OnTimeStateChanged() is calling into
> the VideoRendererSink or we can deadlock the compositor thread.
>
> With video-only playback occurs a buffering state change for underflow
> would immediately invoke OnTimeStateChanged(false) which would try to
> stop the VideoRendererSink under lock.
>
> BUG=496466
> TEST=video only playback doesn't deadlock.
>
> Committed: https://crrev.com/9d8a53f3bf01aed28abbd1f9a713a25362c0d158
> Cr-Commit-Position: refs/heads/master@{#332772}

TBR=xhwang@chromium.org,dalecurtis@chromium.org
NOPRESUBMIT=true
NOTREECHECKS=true
NOTRY=true
BUG=496466

Review URL: https://codereview.chromium.org/1159573005

Cr-Commit-Position: refs/heads/master@{#332808}

media/renderers/video_renderer_impl.cc

@@ -171,12 +171,8 @@ void VideoRendererImpl::Initialize(
   // failed.
   init_cb_ = BindToCurrentLoop(init_cb);
 
-  // Always post |buffering_state_cb_| because it may otherwise invoke reentrant
-  // calls to OnTimeStateChanged() under lock, which can deadlock the compositor
-  // and media threads.
-  buffering_state_cb_ = BindToCurrentLoop(buffering_state_cb);
-
   statistics_cb_ = statistics_cb;
+  buffering_state_cb_ = buffering_state_cb;
   paint_cb_ = base::Bind(&VideoRendererSink::PaintFrameUsingOldRenderingPath,
                          base::Unretained(sink_));
   ended_cb_ = ended_cb;