Fix tokenization bugs and typos

[kingusiu]

Summary

Fixes a handful of bugs encountered while running the VQ-VAE tokenization pipeline.

Changes

• Fix typo in feature dict override: feature_dict_kin_massles_without_cuts.yaml → feature_dict_kin_massless_without_cuts.yaml
• Remove alpha and k optimizer params from the tokenization transformer experiment config — these aren't accepted by RangerLite
• Remove the stray binning_cfg argument passed to tokenize_jetclass_file(...) in create_tokenized_jetclass_files.py, which doesn't accept that parameter and caused a TypeError

Testing

Ran scripts/create_tokenized_jetclass_files.py end-to-end against a trained VQ-VAE checkpoint to confirm tokenization completes without error.