Should Fix #3672

Sorry for last commit which was a little retarded

Signed-off-by: Ilkka Ollakka <ileoo@videolan.org>

This close #2383
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Seems that maxrate needs somekinda exponential function from bitrate (comparing
ffmpeg's libvpx presets) and cbr isn't something you likely will get out of this.

This reverts commit a2770977.

They are only ever used if there is an explicit dst

 This avoids the potential deadlock introduced by
[38ed2e8e] when it is called inside
an event emited by the vout.

This reverts commit 38ed2e8e.
It is not complete.

5cc3984f was too strict, and for baseline it
should be only cosmetic change.

libx264 changed default to normal in that build too, also change logic
littebit so b-pyramid selection works better.

It avoids a vlc_object_detach.

It will be needed if we want to remove one of the timeout in the main vout
thread. It also avoids 1 access of vout->p->p_spu outside of the vout thread.

 The high level object must at least live as long as the input_resource,
I used playlist, vlm or libvlc_media_player.
 It avoids invalid reparent of the vout, aout and sout on input changes.

 It breaks support for per input options used for creating vout, aout
and sout (:sout= is not impacted). I will fix the vout case.

 It breaks invalid usages of vlc_object_find(INPUT, PARENT) used
at least by :
 - the sout modules describe and display.
 - the sout statistics.
 - the vout splitter change
 - the snapshot vout display.
I will fix the vout related ones.

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>