1) remove TBL support in PPC performance. It's much more useful to use the

    PMCs, and with Apple's CHUD it's fairly easy too. No reason to keep useless
    code around
2) make the PPC perf stuff a configure option
3) make put_pixels16_altivec a bit faster by unrolling the loop by 4
patch by (Romain Dolbeau <dolbeau at irisa dot fr>)

Originally committed as revision 2022 to svn://svn.ffmpeg.org/ffmpeg/trunk