Retention-first Bayesian SCR responsivity

estimate_scr_responsivity_bayes() treats SCR responsivity as graded evidence rather than a binary preprocessing deletion rule. It keeps the conventional median-amplitude non-responder flag, estimates a Beta-Binomial posterior response probability, and retains every participant with finite trial data for model-based sensitivity analyses.

This supports a primary-model/robustness workflow in which hard exclusion can be compared with hierarchical retention. The implementation is deliberately transparent and dependency-light; it does not claim to reproduce a Dirichlet-process mixture model.

Methodological motivation includes Thomas & Rabinak (2026), DOI 10.1016/j.biopsycho.2026.109336.