Windows benchmark results are probably wrong

[johnmcfarlane]

At least on appveyor they are. This probably means that results are not correctly escaped so fixing this will also fix this.

[johnmcfarlane]

Google Benchmark DoNotOptimize was added a while back and the numbers look a lot healthier now.

Now:

...
Run on (2 X 2397 MHz CPU s)
CPU Caches:
  L1 Data 32 KiB (x2)
  L1 Instruction 32 KiB (x2)
  L2 Unified 256 KiB (x2)
  L3 Unified 30720 KiB (x1)
-----------------------------------------------------------------------------------
Benchmark                                         Time             CPU   Iterations
-----------------------------------------------------------------------------------
add<float>                                     6.33 ns         6.42 ns    112000000
add<double>                                    6.15 ns         6.14 ns    112000000
add<long double>                               6.13 ns         6.14 ns    112000000
add<u4_4>                                      5.62 ns         5.58 ns    112000000
add<s3_4>                                      5.61 ns         5.63 ns    100000000
add<u8_8>                                      4.94 ns         4.88 ns    112000000
add<s7_8>                                      5.84 ns         5.86 ns    112000000
add<u16_16>                                    5.02 ns         5.02 ns    112000000
add<s15_16>                                    5.90 ns         5.86 ns    112000000
add<int8_t>                                    5.98 ns         6.00 ns    112000000
add<uint8_t>                                   6.00 ns         6.09 ns    100000000
add<int16_t>                                   6.23 ns         6.28 ns    112000000
add<uint16_t>                                  5.29 ns         5.30 ns    112000000
...

Before

...
Run on (2 X 2600 MHz CPU s)
06/10/18 15:38:52
--------------------------------------------------------------------------------
Benchmark                                         Time           CPU Iterations
--------------------------------------------------------------------------------
add<float>                                        2 ns          2 ns  320000000
add<double>                                       2 ns          2 ns  344615385
add<long double>                                  2 ns          2 ns  344615385
add<u4_4>                                         2 ns          2 ns  298666667
add<s3_4>                                         2 ns          2 ns  320000000
add<u8_8>                                         2 ns          2 ns  320000000
add<s7_8>                                         2 ns          2 ns  298666667
add<u16_16>                                       2 ns          2 ns  320000000
add<s15_16>                                       2 ns          2 ns  298666667
add<int8_t>                                       2 ns          2 ns  320000000
add<uint8_t>                                      2 ns          2 ns  298666667
add<int16_t>                                      2 ns          2 ns  320000000
add<uint16_t>                                     2 ns          2 ns  320000000
...