main thread (one more step to hopefully fix #45)
 macosx/controls.m: fixed a tiny memleak

* include/vlc_config.h: VOUT_BOGUS_DELAY was way too low, which triggered continuous trashing of early pictures when seeking in some asf files.

* GET_TIME is inaccurate, because it is based on average block duration.
* In case skipping to DVD_MENU_Title fails try DVD_MENU_Root. Otherwise we might go directly to the movie instead of the menu, which I find less desirable then viewing the FBI warnings.

   priority thread, since these functions use an internal mutex and may
   delay the thread for up to 100 ms under heavy loads (experienced with
   the mosaic). Instead we now recycle the UDP buffers, which is
   incidentally faster than malloc()/free(). It should avoid "late
   buffer sent" messages. Also removed the --sout-udp-late option because
   "better late than never" (I changed my mind on this).

   Allows to run video filter2 modules just before the encoding (and
   after the SPU rendering). New --sout-transcode-high-priority option
   to run the optional encoder thread at the same priority as the output.

* Tell the auhal which channel ordering VLC uses. 7 and 8 channel configs are not supported because VLC uses the channel ordering L R Lm Rm Lr Rr C LFE which isn't a normal ITU, DVD, MPEG, AAC or WAV standard and therefore not supported by default by OS X.

New VLC_EXPORTs are now added at the end of module_symbols_t so that
offsets of other function pointers are preserved.

Deleted VLC_EXPORTs are replaced with dummy pointers for padding.

Of course, it doesn't solve the problem with changing shared structures
format :/ and it may cause SVN conflicts when multiple people changes
VLC_EXPORT. I doubt it can be avoided though.

* Fix several problems with the auhal module. only the problem with the changing inNumberFrames should be left. We need to add a buffer system to take care of this issue...

* Revert the coreaudio resampler to it's simpler variant again. the builtin resampler of the new auhal module takes care of the special case in which this resampler wasn't working. It would be better to really fix it, but i cannot find the problem.