fn std_normal_quantile(p: f64) -> f64
Inverse standard normal CDF (quantile function).
Uses the rational approximation from Peter Acklam (2003). Accurate to roughly 1.15e-9 in the central region.