It fixes #3522.

Close #3520

Close #3513

Signed-off-by: Christophe Mutricy <xtophe@videolan.org>

Sends NULL-pict to encode when it's closing time, then encoder knows to
flush buffers and check that it has outputted all the frames.

Contains changes to x264/avcodec-module to implement that on encoder-side.
Add quick check on omxil/dirac/theora for that, so they don't crash.
And add likely-macro to x264/avcodec in check of NULL

If someone more familiar with dirac/theora/omxil encoder-modules could
check if they have buffers that need to be outputted, would be nice.

Fixes some tickets, but I failed to find any of those in trac.

Multiple cores are pretty common nowdays, so shouldn't be big issue. Maybe
better would be implement cpu-core counter, like in x264, but at the moment
only avcodec-module doesn't seem to handle i_threads=0 as automatic.

Add now needed -lws2 to stream_out_standard

It closes #2812, #3197 and #3533.

They weren't thread-safe (and did not support IPv6).

(I doubt VLC 1.2.0 will be out before July 13th this year)

(cherry picked from commit d16e5aea59a43b335e9414c38900ff57f1fcc562)
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

This reverts commit 5cf83d9c.

It's nor displayed strings but passed to the core