Reland of Injectable Obj-C video codecs

webrtc/sdk/objc/Framework/Classes/VideoToolbox/encoder.mm

Index: webrtc/sdk/objc/Framework/Classes/VideoToolbox/encoder.mm
diff --git a/webrtc/sdk/objc/Framework/Classes/VideoToolbox/encoder.mm b/webrtc/sdk/objc/Framework/Classes/VideoToolbox/encoder.mm
index d3f5ef9a93300eb5ad6b722e0e71c54adf95614b..3674d1869d85d87f7d719c689d97dd153de68487 100644
--- a/webrtc/sdk/objc/Framework/Classes/VideoToolbox/encoder.mm
+++ b/webrtc/sdk/objc/Framework/Classes/VideoToolbox/encoder.mm
@@ -537,9 +537,9 @@ int H264VideoToolboxEncoder::Release() {
   // Need to reset so that the session is invalidated and won't use the
   // callback anymore. Do not remove callback until the session is invalidated
   // since async encoder callbacks can occur until invalidation.
-  int ret = ResetCompressionSession();
+  DestroyCompressionSession();

[jtt_webrtc, 2017/07/20 02:33:20]

This is not the problem. See
ObjCVideoEncoder::~ObjCVideoEncoder()
and
ObjCVideoDecoder::~ObjCVideoDecoder()

[kthelgason, 2017/07/20 11:33:51]

Agreed. I still think we should make this change. This is how it's done in the
decoder, and it makes little sense to recreate the compression session and then
set the callback to null.

[jtt_webrtc, 2017/07/20 17:55:59]

I'm not sure what implications this will have because Reset reinitializes the
capture session and doing this introduces uncertainty.
I rather deal with known issues for now.

   callback_ = nullptr;
-  return ret;
+  return WEBRTC_VIDEO_CODEC_OK;
 }
 
 int H264VideoToolboxEncoder::ResetCompressionSession() {