test: lock dcurves parity for issue #127

[uriahf]

Adds regression coverage for issue #127 without changing implementation.

• compares rtichoke time-dependent DCA with dcurves on the bundled df_surv example at 1.5 years
• checks the reported ~0.22 whole-cohort event prevalence
• compares threshold-specific TP rate, FP rate, and net benefit at representative cutoffs
• adds an exact-threshold boundary case to preserve the prediction >= threshold convention

If these tests pass on current main, #127 can be treated as already resolved by the later pooled-by-cutoff changes.

[github-actions]

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-08-20 07:07 UTC