Add convert_dataset_hf to CLI

[KuuCi]

This PR allows users to call llmfoundry convert_dataset_hf {ARGS} while maintaining correctness with existing convert_dataset_hf script. The motivation is for DLE where we want to make the CLI much more intuitive in the docker images

[KuuCi]

mcli logs test-data-hf-ynEM4L runs:
python data_prep/convert_dataset_hf.py --dataset c4 --data_subset en --out_root my-copy-c4 --splits train_small val_small --concat_tokens 2048 --tokenizer EleutherAI/gpt-neox-20b --eos_text '<|endoftext|>' --compression zstd

mcli logs test-data-hf-253z0P runs:
llmfoundry convert_dataset_hf --dataset c4 --data_subset en --out_root my-copy-c4 --splits train_small,val_small --concat_tokens 2048 --tokenizer EleutherAI/gpt-neox-20b --eos_text '<|endoftext|>' --compression zstd

[irenedea]

Requesting changes around typer usage and naming

[KuuCi]

mcli logs test-data-txt-MYSMm1 runs:
` python convert_dataset_hf.py
--dataset c4 --data_subset en
--out_root my-copy-c4 --splits train_small val_small
--concat_tokens 2048 --tokenizer EleutherAI/gpt-neox-20b --eos_text '<|endoftext|>'
--compression zstd

mcli logs hf-cli-vdGO9f runs:
llmfoundry convert_dataset_hf \ --dataset c4 --data-subset en \ --out-root my-copy-c4 --splits train_small,val_small \ --concat-tokens 2048 --tokenizer EleutherAI/gpt-neox-20b --eos-text '<|endoftext|>' \ --compression zstd

[KuuCi]

mcli logs hf-cli-j8u5rB runs:
llmfoundry data_prep convert_dataset_hf \ --dataset c4 --data-subset en \ --out-root my-copy-c4 --splits train_small,val_small \ --concat-tokens 2048 --tokenizer EleutherAI/gpt-neox-20b --eos-text '<|endoftext|>' \ --compression zstd

mcli logs hf-orig-TGBbOl runs:
python convert_dataset_hf.py \ --dataset c4 --data_subset en \ --out_root my-copy-c4 --splits train_small val_small \ --concat_tokens 2048 --tokenizer EleutherAI/gpt-neox-20b --eos_text '<|endoftext|>' \ --compression zstd

stale