Add unsloth to conda_env. Unfortunately, requires older version of trl => minor python code change needed.

Remove prepare_model_for_kbit_training(...) and get_peft_model(...), because SFTTrainer already cares for it.