Fix NameError: make dataset module self-contained with its own CharTokenizer\n\nDuplicates the minimal CharTokenizer in the dataset module to avoid import dependency issues."
Fix NameError: Add CharTokenizer import to dataset module\n\nAdds 'from neurolex_v4_model import CharTokenizer' after existing imports.\nThis fixes the NameError that occurs when NeuroLexDataset tries to instantiate CharTokenizer."