chromas converstion: optimize MMX/SSE acceleration even more by leveraging out of order execution, basically avoid CPU stalls by using more registers and by spacing memory fetch and use as much as possible (CPU will run available instructions while waiting for memory fecth to complete)

	Use decoder_GetInputAttachment to get attachment from a decoder module.
    You need to release the memory with vlc_input_attachment_Delete.
	Untested...

modules/playlist/* cosmetic.

- video_chroma: added support for IA-32/64 SSE2 accelaration (128 bit vector integer registers), needs LOADS of testing

Qt4 - SysTrayIcon: change the previous commit after some remarks on IRC and fix the toggle/Show-Hide - bug.

chapter encoding, I fear the worst)

 - time/duration display
 - meta info parsing.
 - seek (precise seek if SEEKTABLE presents)
Resampling can still happen but I think it has to do with the
decoder(at least with our ffmpeg wrapper, I haven't tested the
native one).