this patch features :
- persistent resizable main windows
- video confined to preexisting window size
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

The object was typically already killed, so the thread was leaked.
Also, remove vlc_object_kill, since the thread never checks for the
detach of its own object.

This can really only output iamges at this point.
Critical yet-to-be-implemented features:
 - MIT-SHM shared memory support
 - XVideo hardware scaling & chroma conversion
 - keyboard event handling
 - mouse event handling
 - window (manager) event handling
 - cursor blanking
 - window provider submodule (so it can work without the Qt interface)
 - screen capture access module
 - GLX, perhaps?
 - etc

15, 16 and 24-bits color depth work. 32-bits depth was not tested due
to lack of suitable video adapter. 8-bits depth does not work due to a
bug in VLC chroma handling. B&W, grey and 16 color are not supported,
but are probably not worth it - the Xlib output does not support them
either.

Fullscreen should work, but was not tested.

It can only be used on detached ressource to ensure the value validity.

It also probably fixes deadlock between ressources freeing (vout) and input
through gui.

It is very limited:
 - Uncrypted only
 - No ISO or direct disc device
 - No menu
 - No still frame
 - No angle
 - No out of mux stream (some subtitles use it)

It does:
 - title (MPLS)
 - chapter (mark)
 - correct seek
 - hide invalid ES for a given MPLS

So that we don't break the command line of 0.001% of our users

 when checking if fullscreencontrollers position needs to be
 recalculated. Should fix #2283 (what I understand that ticket)

 Please review and test.

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>