GSbench 0.1.0

First release.

• Core (base R): simulate_population(), qc_markers(), impute_markers(), Gmatrix() (VanRaden), and gblup() (GBLUP by REML, validated against rrBLUP::mixed.solve).
• Unified modelling interface gs_fit() / predict() covering GBLUP, elastic net (glmnet), random forest (ranger) and gradient boosting (xgboost).
• gs_cv() for breeding-relevant cross-validation (random k-fold and leave-one-group-out).
• gs_ensemble(), a stacked super-learner combining base models with non-negative, out-of-fold-fitted weights.
• gs_benchmark() with print/summary/plot to compare all available models under one cross-validation.