chestR 0.1.0

• Initial packaged release of kernel-weighted Cox regression (chestr) for exploring treatment effect heterogeneity on a biomarker grid.
• Implements Kish effective-sample-size (ESS) events-per-df safeguard (min_events_per_df, default 10). Grid points below the threshold are skipped with NA coefficients and flagged via reliable.
• Exported API: chestr(), plot.chestr() / plot(), chestr_test().
• Weighting methods: "distance" (default), "square_distance", "legacy".
• Global permutation test of constant treatment effect over biomarker space (T_L2, T_MAX) via chestr_test().
• Vignette chestr-workflow demonstrates the global Cox → chestr → plot workflow (and optional permutation test) on simulated survival data.