Get an Audience validation with scores (experimental)

Experimental. Returns stored validation scores. Older results without question ranges derive them once from the original saved answers and store them; saved scores are not changed. scores.headline is macro MAE in percentage points with a 95% band (clustered bootstrap over Minds and survey sampling, widened by any re-weighting to the Audience's composition) and the noise floor (the MAE sampling alone would produce). scores.kpis lists every KPI of the experiments methodology catalog, each with its value, a 95% interval and the number of questions it covers: MAE, RMSE, total variation distance, Jensen-Shannon divergence and distance, ordinal Wasserstein distance, expected set-size error, top-choice agreement, Pearson and Spearman correlation within questions and pooled across options, and the expected Brier and log scores. scores.validity counts invalid answers by reason (unknown option, too many selections, an exclusive option combined with others, empty); they stay in the denominator. Each question lists human and synthetic shares per option and its own 95% match-score range, using the same sampling and measured population-shift method as the headline. When a scored question has no stated respondent count (baseSource: assumed), its point comparison and existing weight are retained, but its range and noise floor are null. Headline and combined ranges, band bounds and level, and affected KPI intervals are also null when their contributors include such a question. This applies to saved results as well as new scores; a legacy placeholder n is never treated as observed sample size.