Should fix the ppc crash.

Sorry I didn't meant to commit that...

This reverts commit c42248cf.

The changes to the nuv demuxer are:
 - aspect ratio is now used
 - only key frames are used for seeking / index table
 - if there is a seektable it is now used (in a seekable stream)
 - GET_LENGTH is set correctly when a seektable is available, if there
isn't a seektable it will make a guess based on the position in the
stream and the stream size
 - seek support improved, it checks if the stream is seekable and it
also checks if the stream is not changing in size (recording)
 - support for rtjpeg decoder

The demuxer was tested on Windows and Linux with a samba share with
recordings from a mythbackend. I have also tested this with MythWeb
with HTTP Direct Download and ASX Stream. If it is detected that the
stream is not seekable (for example the HTTP stream) it will fall back
to the way it worked before and keep fetching new frames until the
position is reached. If you seek to a position before the current
position it will now give a warning.

If a recording is in progress the stream size will change. This will
be detected and then only forward seeking is possible, because
changing the stream pointer on the recorded file failed (at least it
did using the network share).

Assert rather than invoke a crashy work-around.
Also allow the type-specific cleanup function to join the thread.

Feel free to backport, I don't have the branch yet..

macosx: Do NOT call config_SaveConfigFile() if you don't intend to save ALL the options that might have changed. If you want to autosave something, the config option needs to be marked with change_autosave() and the core will save the option.

The Capture panel is sensibly slower than all the other ones and way less used by most people. Therefore, don't build it (especially for windows) unless you need it. Better for speed and memory.

This reverts commit 9b868a2f.

Also set a correct title and icon, when you don't use the pl through THEDP.

This can happen if you toggle between qt-display-mode too often.

Fix by J-Peg.

Fix bug reported on IRC and on Forum.
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>