fix: restore local execution compatibility and fix setup.py package mapping

perf: subsample to 1000 rows on HF and fix encoder consistency for correct predictions

fix: pre-download SAP model weights and unify HF login

fix: handle inhomogeneous fold counts in statistical analysis

feat: finalized stable architecture with consolidated structure and single-worker mode

perf: reduce CV folds to 3 for faster benchmarking

perf: cache SAP model and fix main.py imports/sys.path

fix: downgrade tabpfn to 0.1.11, class-level weight cache, fix ravel on list

fix: pre-accept TabPFN license in Dockerfile ENV + cache file + all env var names

fix: pad single-row input for SAP RPT-1 OSS predict (np.concatenate crash)

fix: add missing scipy dep, TabPFN weight caching, better error logging

Initial deployment of ModelMatrix-HF