fix: metrics/dataset

fix: dropdown and frontend

fix: made download background and ui sub menu for model select

fix: memory overload by adding top models and options to load singular models

fix: frontend metrics & added model selection for recommendations

fix:v3

refactor: fix the docs and dynamic model load

fix: correct v3 feature order (FEAT_SCALAR+6), fix 0 SOH predictions

fix: predict_array zero-pads X to model n_features_in_, normalize versions to registry

fix: remaining v3 bugs - TFT n_feat detection, simulate URL, features from artifacts

fix: proper v3 implementation - auto-detect n_feat, fix simulate URL, fix model_count

feat: v3 models - XGBoost R2=0.9866, GradientBoosting R2=0.9860 as default

feat: HF model repo at v1/v2 root; version selector UI with download submenu; artifacts/ out of git

feat: full project — ML simulation, dashboard UI, models on HF Hub