Fix per-run combine time (fixes #181)

[patrickdoc]

Before change:

benchmarking perRun
measurement took 231495 s (60 hours!!!)
analysing with 1000 resamples
bootstrapping with 13 of 22 samples (59%)
time                 3.112 ms   (3.102 ms .. 3.122 ms)
                     1.000 R²   (1.000 R² .. 1.000 R²)
mean                 3.119 ms   (3.111 ms .. 3.122 ms)
std dev              10.00 μs   (5.617 μs .. 16.74 μs)
found 1 outliers among 13 samples (7.7%)
  1 (7.7%) low severe
variance introduced by outliers: 4% (slightly inflated)

After:

benchmarking perRun
measurement took 5.135 s
analysing with 1000 resamples
bootstrapping with 33 of 42 samples (78%)
time                 3.120 ms   (3.116 ms .. 3.124 ms)
                     1.000 R²   (1.000 R² .. 1.000 R²)
mean                 3.117 ms   (3.111 ms .. 3.120 ms)
std dev              11.95 μs   (6.974 μs .. 20.56 μs)
found 3 outliers among 33 samples (9.1%)
  3 (9.1%) low severe
variance introduced by outliers: 2% (slightly inflated)

The reported time is now much more accurate, and we get double the number samples.

[RyanGlScott]

Thanks!

As always, make sure to add an entry to the changelog describing what's different.

[RyanGlScott]

It might also be wise to add a comment describing why we add the Measureds together, but don't add the times together.

[patrickdoc]

Done and done. I promise I will remember the changelog in the future 😣