test packed dataset with preference dataset

[felipemello1]

preference datasets, like stack exchange havent been tested with dataset.packed=True yet.

We should run it for a full epoch, make sure it works, and update it to accept packed=True

[SalmanMohammadi]

    sys.exit(recipe_main(conf))
             ^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/dist-packages/recipes/lora_dpo_single_device.py", line 620, in recipe_main
    recipe.setup(cfg=cfg)
  File "/usr/local/lib/python3.11/dist-packages/recipes/lora_dpo_single_device.py", line 219, in setup
    self._sampler, self._dataloader = self._setup_data(
                                      ^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/dist-packages/recipes/lora_dpo_single_device.py", line 359, in _setup_data
    ds = config.instantiate(cfg_dataset, tokenizer=self._tokenizer)
         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/dist-packages/torchtune/config/_instantiate.py", line 106, in instantiate
    return _instantiate_node(OmegaConf.to_object(config), *args)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/dist-packages/torchtune/config/_instantiate.py", line 31, in _instantiate_node
    return _create_component(_component_, args, kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/dist-packages/torchtune/config/_instantiate.py", line 20, in _create_component
    return _component_(*args, **kwargs)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/dist-packages/torchtune/datasets/_stack_exchange_paired.py", line 141, in stack_exchange_paired_dataset
    return PackedDataset(ds, max_seq_len=tokenizer.max_seq_len)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.11/dist-packages/torchtune/datasets/_packed.py", line 100, in __init__
    self._pack()
  File "/usr/local/lib/python3.11/dist-packages/torchtune/datasets/_packed.py", line 124, in _pack
    tokens, labels = sample["tokens"], sample["labels"]

because preference dataset returns

        tokenized_dict = dict(
            chosen_input_ids=chosen_input_ids,
            chosen_labels=chosen_labels,
            rejected_input_ids=rejected_input_ids,
            rejected_labels=rejected_labels,
        )