The dashboard (tjbench/dashboard.py) renders point estimates (pass rates) but not the confidence interval
around them, which is the honest way to read a small-sample result. Surface the Wilson interval per suite in
the dashboard tables so a viewer sees the error bar, not just the number.
Acceptance criteria: per-suite pass rates render with their CI in the dashboard; offline test covers the
rendering; no point estimate is shown without its interval.
The dashboard (
tjbench/dashboard.py) renders point estimates (pass rates) but not the confidence intervalaround them, which is the honest way to read a small-sample result. Surface the Wilson interval per suite in
the dashboard tables so a viewer sees the error bar, not just the number.
Acceptance criteria: per-suite pass rates render with their CI in the dashboard; offline test covers the
rendering; no point estimate is shown without its interval.