docs: add IVAC serosurvey demo notebook

[jc-macdonald]

Add an executed 15-minute notebook for a mixed IVAC audience comparing 18 synthetic serosurvey designs by financial cost, overall prevalence-estimation error and underserved-group error. The narrative connects to IVAC's serosurveillance/costing work, exposes the Study and preference API, and includes four figures, result tables, presenter notes and budget/priority controls that reuse existing results.

The companion model preserves population weights under oversampling, exact participant totals and independent phase/design/replicate streams. Both phases evaluate all designs; refinement changes simulation replication while preserving the population assumptions. All population inputs, prices and preference weights are explicitly hypothetical; Monte Carlo precision is distinguished from survey and model uncertainty.

Include a small notebook extra, an illustrated guide, saved notebook outputs and CI execution in a fresh kernel with a 60-second per-cell timeout. Regression tests cover population weighting, known clustered-estimator mean/variance, reproducibility, financial costing and budget/no-choice behavior.

Validation: just ci (605 tests, 97.68% coverage), strict documentation build, notebook format/lint, isolated execution using only the notebook extra, and repeated fresh-kernel execution in 6–8 seconds including rendering. The companion script evaluates 19,800 surveys and saves four figures in about two seconds. Saved outputs were exported to HTML as a presentation backup.