[Benchmarks] Throw usage error when using dataset-name random and dataset-path together

[yeqcharlotte]

Purpose

There's some confusion #24684 that default behavior for vllm bench serve is incorrect. This is due to vllm bench serve uses random by default and it does not throw any errors when using provide a shardgpt dataset path lol.

Test Plan

# only specify dataset path
vllm bench serve --model meta-llama/Llama-3.1-8B-Instruct --percentile-metrics ttft,tpot,itl,e2el --dataset-path /models/ShareGPT_Vicuna_unfiltered/ShareGPT_V3_unfiltered_cleaned_split.json --max-concurrency 64 --num-prompts 640 --ignore-eos --sharegpt-output-len 512
# specify dataset path first, then dataset name
vllm bench serve --model meta-llama/Llama-3.1-8B-Instruct --percentile-metrics ttft,tpot,itl,e2el --dataset-path /models/ShareGPT_Vicuna_unfiltered/ShareGPT_V3_unfiltered_cleaned_split.json --dataset-name random --max-concurrency 64 --num-prompts 640 --ignore-eos --sharegpt-output-len 512
# specify dataset name first, then dataset path
vllm bench serve --model meta-llama/Llama-3.1-8B-Instruct --dataset-name random --percentile-metrics ttft,tpot,itl,e2el --dataset-path /models/ShareGPT_Vicuna_unfiltered/ShareGPT_V3_unfiltered_cleaned_split.json --max-concurrency 64 --num-prompts 640 --ignore-eos --sharegpt-output-len 512

Test Result

all throw the following error:

usage: vllm bench serve [options]
vllm bench <bench_type> [options] serve: error: Cannot use 'random' dataset with --dataset-path. Please specify the appropriate --dataset-name (e.g., 'sharegpt', 'custom', 'sonnet') for your dataset file: /models/ShareGPT_Vicuna_unfiltered/ShareGPT_V3_unfiltered_cleaned_split.json

---

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

[houseroad]

Looks good.

[houseroad]

nit: also dataset_path is not empty string, like ""?