POWER cycle count numbers are wrong

[randombit]

Eg from #1459 on a POWER8 machine

AES-128 encrypt buffer size 1024 bytes: 1952.789 MiB/sec 0.25 cycles/byte (5858.37 MiB in 3000.00 ms)

The hardware AES maxes out at ~ .7 cycles/byte (and #1459 is not well optimized, it doesn't even pipeline the calls) so .25 cycles/byte is totally bogus.

[randombit]

Two problems, we are in fact calling the cycle counter incorrectly (failing to account for upper timebase overflowing) but also POWER8s cycle counter runs at only 512 MHz.

[randombit]

We need a way to specify the cycle counter rate to speed util for reporting.

[randombit]

Fixed