convert vector_fmul_reverse_sse2 and vector_fmul_add_add_sse2 to sse

please complain if they are slower on sse2 cpus ...

Originally committed as revision 5976 to svn://svn.ffmpeg.org/ffmpeg/trunk