Initialize the alpha plane pointers even when not used !

This partly fixes #2815, but the conflict of "Reveal in Finder" and "Record" has not yet been fixed.

When using the hotkeys to seek, each offset seek was executed twice. The bahavior of INPUT_CONTROL_SET_TIME_OFFSET has changed, and these callbacks did not account for that yet.

This closes #2820

Just to get compilation working. Any use of Timer API will result in an abort() for now.

Idea and code from Francois Cartegnie

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

- video.deinterlaceEnable(char *mode) : enable deinterlace filter which type is defined by mode
     - video.deinterlaceDisable()          : disable deinterlace filter
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

- video.deinterlaceEnable(char *mode) : enable deinterlace filter which type is defined by mode
 - video.deinterlaceDisable()          : disable deinterlace filter
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

Patch by ozvald (see #2733)

  - libvlc_video_enable_deinterlace(libvlc_media_player_t *p_mi, int b_enable, const char *psz_mode, libvlc_exception_t *p_e)
      - b_enable: boolean to enable or disable deinterlace filter
      - psz_mode: char to define the deinterlace mode (blend, linear...)
Signed-off-by: Laurent Aimar <fenrir@videolan.org>

This string is already in use, so no problem with stringfreeze for
1.0 (According to xtophe).

In general, we should avoid PL_LOCK'ing around stuff that the Datasource does,
and restrict it to the lines that directly deal with the core playlist.

This should fix #2810

This closes #2814

It allows to use mouse inside a simple "video filter2"
It is not meant to stay.

Unfortunately, the thread stack remains allocated as our threading
abstraction does not support detaching.

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>