Add optional warmup test suite

[camillobruni]

We've noticed that we get lazy compilation events for test-runner methods.
By doing running a warmup suite first, we can avoid this addition noise during measurements.

• Add Warmup suite
• Add better performance.measure annotation for all suites

[rniwa]

I don't think we want to do this. We want to measure junks caused by compilation of test harness code since it's a real cost.

[bgrins]

I don't think we want to do this. We want to measure junks caused by compilation of test harness code since it's a real cost.

I don't know - it seems like a somewhat random cost to include in the measurements (both because it only applies on the first tests using these functions, and because "benchmark runner" isn't a particularly representative workload that we'd include on its own merits).

If this effectively reduces noise in measurements I think we should consider it. @camillobruni how does this affect the scores and the deviation across runs?

[camillobruni]

Ro exaggerate the effects on warmup, I ran with iterationCount=3 averaged over 10 runs:

Without warmup:

browser                               Google Chrome Dev  Safari
version                               114.0.5696.0       16.4
-------------------------------------------------------------------
Editor-CodeMirror/total               15.58 ± 3.7%       21.7 ± 7.1%
Editor-TipTap/total                   11.53 ± 3.7%       156 ± 40%
React-Stockcharts-SVG/total           111.7 ± 4.3%       148.0 ± 6.0%
React-Stockcharts/total               86.8 ± 4.7%        121.8 ± 4.6%
TodoMVC-Angular/total                 21.9 ± 5.4%        27.1 ± 4.6%
TodoMVC-Backbone/total                25.07 ± 2.6%       32.8 ± 5.1%
TodoMVC-JavaScript-ES5/total          34.16 ± 2.5%       30.2 ± 3.7%
TodoMVC-JavaScript-ES6-Webpack/total  39.2 ± 3.5%        35.8 ± 3.8%
TodoMVC-JavaScript-ES6/total          43.0 ± 2.6%        39.6 ± 3.2%
TodoMVC-Preact/total                  9.13 ± 3.0%        12.08 ± 5.4%
TodoMVC-React-Redux/total             37.8 ± 2.7%        44.0 ± 4.9%
TodoMVC-React/total                   31.89 ± 3.1%       38.5 ± 4.3%
TodoMVC-Vue/total                     14.98 ± 4.3%       19.0 ± 5.6%
TodoMVC-jQuery/total                  87.2 ± 2.7%        85.0 ± 4.0%
-------------------------------------------------------------------
Score                                 644 ± 2.7%         469 ± 4.6%

With warmup

browser                               Google Chrome Dev  Safari
version                               114.0.5696.0       16.4
-------------------------------------------------------------------
Editor-CodeMirror/total               15.65 ± 1.7%       23.0 ± 5.1%
Editor-TipTap/total                   11.40 ± 2.1%       195 ± 11%
React-Stockcharts-SVG/total           117.40 ± 0.78%     155.1 ± 2.3%
React-Stockcharts/total               89.8 ± 2.2%        126.0 ± 0.81%
TodoMVC-Angular/total                 22.58 ± 1.1%       27.56 ± 1.5%
TodoMVC-Backbone/total                25.62 ± 3.0%       33.55 ± 2.1%
TodoMVC-JavaScript-ES5/total          34.62 ± 1.2%       30.62 ± 1.3%
TodoMVC-JavaScript-ES6-Webpack/total  39.93 ± 1.9%       36.60 ± 1.4%
TodoMVC-JavaScript-ES6/total          43.48 ± 1.3%       40.47 ± 1.8%
TodoMVC-Preact/total                  9.42 ± 3.8%        12.80 ± 4.5%
TodoMVC-React-Redux/total             39.06 ± 1.5%       45.2 ± 2.4%
TodoMVC-React/total                   32.83 ± 2.3%       38.98 ± 2.4%
TodoMVC-Vue/total                     15.54 ± 2.2%       19.38 ± 3.2%
TodoMVC-jQuery/total                  89.29 ± 0.80%      86.2 ± 1.6%
-------------------------------------------------------------------
Score                                 628.5 ± 0.71%      445.1 ± 0.98%

Looks like both browsers have rather slow first runs.
The Score variance is a bit less extreme in reality, since likely the metrics for Editor and StockChart are off.