* added the new OSX vout files to pb proj.

  (it looks for .vlt files in all the directories of the resource path)
  * dialogs.cpp: fixed a stupid copy/paste mistake

 * skins2/parser/interpreter.cpp: Renamed vlc.ontop() into vlc.onTop()
 * doc/skins/skins2-howto.xml: doc updated

  (fixes glx when X11 isn't runnning)

  "Restore Defaults" so it gets translated

* po/fr.po : Chanded "Restore Defaults" translation
  

   doesn't work. Better than nothing but the sound is still poor with DVDs :(

   snd_pcm_status_get_tstamp() returns zero. I hope someone has a better
   fix because in fact it is not perfect (there are "clicks" with DVDs)

   aout_OutputNextBuffer(), then do not try to resample or play silences.
   Now everything works perfectly with screwed ALSA drivers that always
   return zero in snd_pcm_status_get_tstamp() (at last!!!)

   The parsing of the title/chapter section is more strict so shouldn't trigger on false positives like udp syntax or http url with login/password.
* src/libvlc.h: updated useage.

    + Don't use vlc_app_DATA for the BeOS bundle.
    + Build the BeOS bundle in $(top_builddir).

* src/input/input.c: hack for udp access and MRLSections().

  - syntax is: [url][@[title-start][,chapter-start][-[title-end][,chapter-end]]]
  - core also handles start/end boundaries itself (simplifies the access plugins).
* modules/access/dvdread.c,dvdnav.c,cdda.c: removed MRL parsing code.

    + Added macros for AltiVec versions of these modules. Will not work at all
      for the moment.

   (really strange crashes happen otherwise, can't figure out why)
 + Makefile.am: don't forget to build vlc-bundle on BeOS

  - if no title-start/end are specified, the whole cd is played.
  - if only a title-start is specified, only this title is played.
  - if a title-start and end are specified, this range is played.
  - "@title-start-" will play everything from the specified start position.

  though there're not 100% posix compliant - this way we can build vlc
  with dynamic plugins (not linked against libintl) again

api gives proper detection for too small buffer. Thanks to Ross for pointing
this out.