Ignore wholly translucent rows at the top and bottom of the subtitle page

Only create an overlay large enough for the remaining rows and only only ask
zvbi to render those rows.

One happy side effect is that now --vbi-text no longer outputs header text
for some teletext streams (although --vbi-text is still unusable imho).

Reducing the size of the overlay allows stutter free playback of teletext
subtitles on significantly slower hardware.

The code blocks in libva vaSyncSurface() when the decoding context has been destroyed.
The problem is observed with a 'via' vdpau driver.

Use AVPacket API to keep PTS and DTS with the picture frames. Part of
this commit is a manual backport of modification in videolan master repo.

VLC frequently reports a framerate that is twice the actual framerate for interlaced content.
Correctly take the information from ffmpeg in account when determining the framerate. It now
correctly reports the actual framerate of the movie.

Inside vaapi.c the function vlc_va_Release() only has a VASurfaceID when
the ffmpeg Frame buffer is no longer needed. The VA Connection struct works
with vlc_va_surface_t pointer, thus a way is needed to find the surface
which belongs to this VASurfaceID. Initially this was coded inside the
vlc_va_Release() functionality, but that exposes details that should be
remain inside va.c file.

The surface handling has been rewritten to be more consistent and to
avoid reusing surfaces that are still in use by the display thread. Some
VAAPI wrapper drivers use the DirectSurface mode, meaning that the surface
handed by the application is used directly for displaying. It does not
copy the surface into a display surface. When the driver does this, then
you cannot reuse this surface which is being displayed for rendering the
next picture too

Conflict: configure.ac
- Keep this tree on version 2.0.6 on this point.

By default the VLC core tries to recycle video outputs. In case of AMD/ATI
graphics cards (as on the AMD Fusion archtecture) this value should be false.
The default value 'false' also works for NVidia and Intel Graphics cards, thus
it is safe to leave the value as is.

Check also in runtime if the create VAImage succeeds properly.

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>
(cherry picked from commit 2339bb72f92e884904b519098027cce453666bde)

and syncronize with xcode 3 project
close #8382

(not cherry-picked from 4065770e62dfe240c597c4154079a21c6ce3f7fe)

(cherry picked from commit bf02b8dd211d5a52aa301a9a2ff4e73ed8195881)
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>

(cherry picked from commit 32bfb8dc02f755d46761268f5674eb56671c81f5)
Signed-off-by: Rafaël Carré <funman@videolan.org>

It is not needed anymore since ce6521381b , parsing will
stop at first unrecognized character (like \r or \n)
(cherry picked from commit 6841a105dad5bf416fe7e1b906bf4abdce21fbe9)
Signed-off-by: Rafaël Carré <funman@videolan.org>

(cherry picked from commit 3b1ecaf7ae893a53b1d8e4928824e10ca20ac2e9)
Signed-off-by: Rafaël Carré <funman@videolan.org>

Signed-off-by: Christoph Miebach <christoph.miebach@web.de>

Signed-off-by: Christoph Miebach <christoph.miebach@web.de>

Signed-off-by: Christoph Miebach <christoph.miebach@web.de>

Signed-off-by: Christoph Miebach <christoph.miebach@web.de>

Signed-off-by: Christoph Miebach <christoph.miebach@web.de>

Signed-off-by: Christoph Miebach <christoph.miebach@web.de>

Signed-off-by: Christoph Miebach <christoph.miebach@web.de>

Signed-off-by: Christoph Miebach <christoph.miebach@web.de>

Signed-off-by: Christoph Miebach <christoph.miebach@web.de>

Signed-off-by: Christoph Miebach <christoph.miebach@web.de>