Add Luganda dialect-aware tokenizer trained on WaxalNLP + SALT

initial commit