WebAudio: record whether user gesture was available and provided.

third_party/WebKit/Source/modules/webaudio/AudioBufferSourceNode.cpp

Index: third_party/WebKit/Source/modules/webaudio/AudioBufferSourceNode.cpp
diff --git a/third_party/WebKit/Source/modules/webaudio/AudioBufferSourceNode.cpp b/third_party/WebKit/Source/modules/webaudio/AudioBufferSourceNode.cpp
index c1ff657683969e77b6d366583016fa35740e166f..f0d121d5d317cf6e3840181cfddde18e9c43bbe1 100644
--- a/third_party/WebKit/Source/modules/webaudio/AudioBufferSourceNode.cpp
+++ b/third_party/WebKit/Source/modules/webaudio/AudioBufferSourceNode.cpp
@@ -441,6 +441,8 @@ void AudioBufferSourceHandler::startSource(double when, double grainOffset, doub
 {
     ASSERT(isMainThread());
 
+    context()->recordUserGesture();

[Raymond Toy, 2016/06/03 17:57:52]

Not sure if this is the right place.  Say you've created a context and set up a
bunch of sources.  Then you schedule the source to start 1 minute from now. 
You'll get the user gesture recorded, but the sound wouldn't have started
playing for another minute.  

Don't you want to know that when the sound actually starts playing that you
didn't have the user gesture permission to play?

[mlamouri (slow - plz ping), 2016/06/05 14:44:57]

As above, I'm following WebKit's behaviour here. The rationale is that if
start() is called with a gesture, it will unlock the AudioContext which will
then be able to start running when startPlaying() is called. Otherwise, if
`start(1000)` is called with a gesture, when the playback actually starts, there
will be no gesture on the stack. Does that make sense?

[Raymond Toy, 2016/06/06 15:10:07]

Acknowledged.

+
     if (playbackState() != UNSCHEDULED_STATE) {
         exceptionState.throwDOMException(
             InvalidStateError,