fix: remove CNN from build (OOM), inline tabular training, ELA heuristic fallback for image

fix: remove XGBoost use_label_encoder (removed in v2+), reduce CNN to 3k imgs/5 epochs for HF timeout, pre-download ResNet18

feat: hybrid CNN training with real HF cert images + fix Dockerfile offline vars bug

feat: migrate to HF Spaces — ResNet-18 CNN + BERT + DistilBERT + tabular ensemble

feat: optimize for 512MB memory (Render free tier)

FIX: optimize training models

CREATE : Trained the model