(cherry picked from commit c8abfa0c)

(cherry picked from commit 28568f40)

-G711
-bv16
-G729
-speex
-G722
-GSM

(cherry picked from commit 75fde98a)

(cherry picked from commit 36fffb05)

(cherry picked from commit dbe78d5f)

Real time text fix. Handle the case where t140 was the prefered codec selected by SDP offer/answer but finally red is used.

VideoToolboxDecoder: add mechanism that disables the effect of outputCb() while the decoding seesion is destroying

(cherry picked from commit 422b34f2)

The H264 encoder on some Huawei devices doesn't take care of row strides information
provided by MediaImage object. That resulted on bad pixel conversion
while input raw pictures were copied into the encoder input buffer.

The workaround is to assume that the encoder take yuv420p pictures
as input and thus to copy each picture in the input buffer without
any pixel format conversion.

This workaround is only applied on Huawei GRA-L09 and VNS-L31 for now.

It was set a long time ago, because iOS was slow in starting the camera.
This is no longer the case, and doing this in a background thread is unsafe.
Indeed, if the call is terminated rapidly, the start() may happen while the video graph is already destroyed.

* ensure that all the capture filters write ticker timestamp
  in generated mblk_t.
* make vp8 encoder to copy the timestamp of each input buffer in
  corresponding output buffers.
* add a trace in MSRtpSend filter printing the delay between the
  moment the picture was captured by the camera and the moment
  the encoded picture is sent on the network.

If ffmpeg is not compiled in, the video preview in the corner of the screen won't be displayed.

It sometimes take a lot of time, which is painful for test suites.

- Fix when the first TMMBR is inferior to the default size's required bitrate
- Fix encoder's bitrate not updated when in custom preset

Attempt to keep compatibility with buggy openssl fragmenting Client Hello packet
and setting fragments in one DTLS packet