Auto sampling for statistical significance

[neonstalwart]

https://github.com/google/tachometer?tab=readme-ov-file#sampling describes flags that tachometer uses to automatically determine sample size so that performance comparisons are statistically significant. A similar approach could be used to automatically set the sample size for web-codegen-scorer. Use the first n samples (e.g. 50) to estimate current rate of pass/fail and then find the sample size needed for 95% confidence and 5% margin of error for the results. Initial sample size, confidence level, margin of error could all be configurable via flags.