Really, really remove all mention of GstNetBuffer

Fixes https://bugzilla.gnome.org/show_bug.cgi?id=673510

Link pbutils and encoding tests against internal version of libgstvideo.

From 7fda524 to 464fe15

Avoid pushing out buffers with the same timestamp only if the out buffers are
decoded from the same input buffer. Instead keep the timestamps when upstream
pushes consecutive buffers with the same ts.

Set playsink's send-event-mode to MODE_FIRST as playbin2 only
needs one event going to the demuxer for its operation

https://bugzilla.gnome.org/show_bug.cgi?id=673211

Adds a property for playsink to define how it should handle
events sent in send_event function. The default is the same as
GstBin's, sending events to all internal sinks. There is also
mode-first, that will send to sinks until the one handles the
event successfully.

https://bugzilla.gnome.org/show_bug.cgi?id=673211

Ensure writability of rtp buffer and remove old memory first
Fix some docs

... such as setting input caps, ensuring to unmap and bearing in
mind that gst_buffer_make_writable usually does not provide
a separate memory copy/area.

... to really only test the indicated scaling method rather than
all of them.

Deactivate the old bufferpool when we negotiate a new one.

Fixes https://bugzilla.gnome.org/show_bug.cgi?id=673327

https://bugzilla.gnome.org/show_bug.cgi?id=673056

Prepare for the future, make it possible to map multiple buffer regions, like
the header and the payload.

Use _find_memory and _map_range to simplify, improve and optimize the
memory mapping of video frames.