MLP DSP functions x86-optimized.
12.59% overall speedup in x86_32
 9.98% overall speedup in x86_64
compared to gcc 4.3.3

It breaks on x86_64 with PIC because the MANGLE macro appends (%%rip), which
is not wanted for labels.

Originally committed as revision 18904 to svn://svn.ffmpeg.org/ffmpeg/trunk