refactor: split train and val dataset in preference dataset

[yuki-97]

Closes #1050.

1. Split train and val in build-in preference dataset, so that we could unblock multiple dataset support.
2. Unify the built-in datasets under nemo_rl/data/datasets/preference_datasets/ into a similar format.
3. Move setup_preference_data to nemo_rl/data/utils.py and reuse it.

Usage

data:
  # other data settings, see `examples/configs/sft.yaml` for more details
  ...
  # dataset settings
  train:
    # this dataset will override prompt_key and use the default values for other vars
    data_path: /path/to/local/train_dataset.jsonl  # local file or hf_org/hf_dataset_name (HuggingFace)
    prompt_key: context
    split: train  # used for HuggingFace datasets
  validation:
    # this dataset will use the default values for other vars except data_path
    data_path: /path/to/local/val_dataset.jsonl
  default:
    # will use below vars as default values if dataset doesn't specify it
    dataset_name: BinaryPreferenceDataset
    prompt_key: prompt
    chosen_key: chosen
    rejected_key: rejected
    prompt_file: null
    system_prompt_file: null

Migrate Guide

1. For dataset that loads from local JSONL file or HuggingFace (BinaryPreferenceDataset and PreferenceDataset)

# old
data:
  dataset_name: BinaryPreferenceDataset
  train_data_path: <PathToTrainingDataset>  # e.g., /path/to/local/dataset.jsonl or hf_org/hf_dataset_name (HuggingFace)
  val_data_path: <PathToValidationDataset>
  prompt_key: <PromptKey>, default is "prompt"
  chosen_key: <ChosenKey>, default is "chosen"
  rejected_key: <RejectedKey>, default is "rejected"
  train_split: <TrainSplit>, default is None  # used for HuggingFace datasets
  val_split: <ValSplit>, default is None  # used for HuggingFace datasets

# new
data:
  # other data settings, see `examples/configs/sft.yaml` for more details
  ...
  # dataset settings
  train:
    # this dataset will override prompt_key and use the default values for other vars
    data_path: /path/to/local/train_dataset.jsonl  # local file or hf_org/hf_dataset_name (HuggingFace)
    prompt_key: context
    split: train  # used for HuggingFace datasets
  validation:
    # this dataset will use the default values for other vars except data_path
    data_path: /path/to/local/val_dataset.jsonl
  default:
    # will use below vars as default values if dataset doesn't specify it
    dataset_name: BinaryPreferenceDataset
    prompt_key: prompt
    chosen_key: chosen
    rejected_key: rejected
    prompt_file: null
    system_prompt_file: null

2. For some built-in datasets that needs change
   1. HelpSteer3

      # old
      data:
        dataset_name: HelpSteer3
      
      # new
      data:
        train:
          dataset_name: HelpSteer3
          split: train
        validation:
          dataset_name: HelpSteer3
          split: validation

   2. Tulu3Preference

      # old
      data:
        dataset_name: Tulu3Preference
      
      # new
      data:
        train:
          dataset_name: Tulu3Preference
        validation: null

Test Result
Nightly tests are all good.

algo	result
rm
dpo