Fix nvbench output for sha512

[davidwendt]

Description

Fixes the sha512 output for nvbench for GlobalMem BW.
Previously:

|    65536 |     0 | sha512 |   1216x | 417.898 us |  1.40% | 412.669 us | 0.61% |     24.139 GB/s |           3.14% |
| 16777216 |     0 | sha512 |     11x |  71.392 ms |  0.03% |  71.387 ms | 0.03% | 258404.649 PB/s | 33642233417.78% |
|    65536 |   0.1 | sha512 |   1184x | 433.031 us |  1.58% | 427.815 us | 1.01% |     22.919 GB/s |           2.98% |
| 16777216 |   0.1 | sha512 |     11x |  73.457 ms |  0.03% |  73.452 ms | 0.03% | 251140.174 PB/s | 32696456458.71% |

Fixed integer overflow calculation:

|    65536 |     0 | sha512 |   1200x | 423.838 us |  1.42% | 418.561 us | 0.66% |  23.799 GB/s |  3.10% |
| 16777216 |     0 | sha512 |     11x |  72.773 ms |  0.11% |  72.767 ms | 0.11% |  35.041 GB/s |  4.56% |
|    65536 |   0.1 | sha512 |   1168x | 439.078 us |  1.60% | 433.843 us | 1.05% |  22.601 GB/s |  2.94% |
| 16777216 |   0.1 | sha512 |     19x |  75.108 ms |  0.49% |  75.102 ms | 0.49% |  33.412 GB/s |  4.35% |

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[bdice]

Thanks!

(Too bad that we can't achieve that many petabytes per second.)

[davidwendt]

/merge