Fix sepia video filter in PackedYUV

SSE function used in PackedYUV plane was bad, removed
copying data using Memcpy8BMMX was slower than vlc_memset -> changed
Sepia8ySSE4_1 changed to Sepia8ySSE2, due to some changes I made to speed it up and clean
Minor changes in indentation and comments
Signed-off-by: Jean-Baptiste Kempf <jb@videolan.org>