Move the initial glyph cache cost out of the measured part of the perf dashboard workload.

[rniwa]

This makes the benchmark results more consistent when re-running it in the same tab.

[rniwa]

Here are before/after numbers. It basically doubles each engine's score.

Chrome

• Before

  • 17.1 ± 1.9
  • 16.9 ± 1.3
  • 16.8 ± 1.3

• After

  • 31.5 ± 1.7
  • 31.5 ± 1.1
  • 31.3 ± 1.4

Firefox

• Before

  • 5.91 ± 0.32
  • 5.96 ± 0.23
  • 5.93 ± 0.25

• After

  • 9.94 ± 0.19
  • 9.99 ± 0.25
  • 10.1 ± 0.27

Safari

• Before

  • 13.6 ± 2.2
  • 13.8 ± 2.3
  • 14.2 ± 2.3

• After

  • 26.1 ± 0.65
  • 26.7 ± 0.94
  • 26.7 ± 0.95

absorb-initial-cost-in-perf-dashboard.zip