chacha/asm/chacha-ppc.pl: improve POWER8 performance by 15%.

This comes at cost of minor 2.5% regression on G4, which is reasonable
trade-off. [Further improve compliance with ABI requirements.]

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/6406)