Prepare the AudioSendStream to be hooked up to send-side BWE.

talk/media/webrtc/webrtcvoiceengine.cc

Index: talk/media/webrtc/webrtcvoiceengine.cc
diff --git a/talk/media/webrtc/webrtcvoiceengine.cc b/talk/media/webrtc/webrtcvoiceengine.cc
index 4bd0400dfa8e394642f22a65f77a601b7efec2f7..a3fd09aafd0356b35d5ea6f610da14c65e15152e 100644
--- a/talk/media/webrtc/webrtcvoiceengine.cc
+++ b/talk/media/webrtc/webrtcvoiceengine.cc
@@ -537,6 +537,7 @@ void WebRtcVoiceEngine::Construct() {
         kRtpTransportSequenceNumberHeaderExtensionDefaultId));
   }
   options_ = GetDefaultEngineOptions();
+  voe_config_.Set<webrtc::VoicePacing>(new webrtc::VoicePacing(true));

[the sun, 2015/11/30 12:37:20]

If this is always enabled, don't put it in the config.

[stefan-webrtc, 2015/11/30 15:22:02]

It's a question of whether we want to keep supporting the old API or not, since
it isn't possible to set the congestion control objects (and RtpPacketSender in
particular) to the voice channel in the old API (and it doesn't make sense to
try to do it with the old API either).

[the sun, 2015/12/01 10:25:35]

Given the time frame for deprecating the old API isn't astronomical, I think
this can be a feature that's only available by hooking up an AudioSendStream to
a channel.

 }
 
 WebRtcVoiceEngine::~WebRtcVoiceEngine() {