[Bugfix] Use LoadFormat values as choices for vllm serve --load-format

[mgoin]

The choices for the --load-format argument haven't been kept in sync with new load formats being added. This PR changes the choices of the arg to directly reference the LoadFormat enum.

For instance, before this PR you could not load sharded_state models with vllm serve

> huggingface-cli download neuralmagic/Meta-Llama-3-8B-Instruct-FP8 --local-dir Meta-Llama-3-8B-Instruct-FP8
> python examples/save_sharded_state.py --model Meta-Llama-3-8B-Instruct-FP8 --tensor-parallel-size 2 --output Meta-Llama-3-8B-Instruct-FP8-SHARDED
> vllm serve Meta-Llama-3-8B-Instruct-FP8-SHARDED --tensor-parallel-size 2 --load-format sharded_state
usage: vllm serve <model_tag> [options]
vllm serve: error: argument --load-format: invalid choice: 'sharded_state' (choose from 'auto', 'pt', 'safetensors', 'npcache', 'dummy', 'tensorizer', 'bitsandbytes')

This did not accurately list the enum values we have available

class LoadFormat(str, enum.Enum):
    AUTO = "auto"
    PT = "pt"
    SAFETENSORS = "safetensors"
    NPCACHE = "npcache"
    DUMMY = "dummy"
    TENSORIZER = "tensorizer"
    SHARDED_STATE = "sharded_state"
    GGUF = "gguf"
    BITSANDBYTES = "bitsandbytes"

[github-actions]

👋 Hi! Thank you for contributing to the vLLM project.
Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run fastcheck CI which consists a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of default ones by unblocking the steps in your fast-check build on Buildkite UI.

Once the PR is approved and ready to go, please make sure to run full CI as it is required to merge (or just use auto-merge).

To run full CI, you can do one of these:

• Comment /ready on the PR
• Add ready label to the PR
• Enable auto-merge.

🚀

[DarkLight1337]

Thanks for fixing!