[Bugfix] Prevent benchmark_throughput.py from using duplicated random prompts

[mgoin]

Currently this benchmark script is misleading performance for V1 (as in the performance is too good) since prefix caching is enabled by default and all of the prompts in the default case are identical.

Before this PR:

VLLM_USE_V1=1 python benchmarks/benchmark_throughput.py --model meta-llama/Llama-3.1-8B-Instruct --num-prompts 1000 --input-len 128 --ou
tput-len 128 --max-model-len 4096
...
Throughput: 151.99 requests/s, 38910.45 total tokens/s, 19455.23 output tokens/s

After this PR:

VLLM_USE_V1=1 python benchmarks/benchmark_throughput.py --model meta-llama/Llama-3.1-8B-Instruct --num-prompts 1000 --input-len 128 --ou
tput-len 128 --max-model-len 4096
...
Throughput: 95.80 requests/s, 24524.24 total tokens/s, 12262.12 output tokens/s

"Reference" baseline of V0 + --num-scheduler-steps 8:

python benchmarks/benchmark_throughput.py --model meta-llama/Llama-3.1-8B-Instruct --num-prompts 2000 --input-len 128 --output-len 128 --max-model-len 4096 --num-scheduler-steps 8
...
Throughput: 97.71 requests/s, 25014.13 total tokens/s, 12507.06 output tokens/s

[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 starts running only a small and essential subset of CI tests to quickly catch errors. You can run other CI tests on top of those by going to your fastcheck build on Buildkite UI (linked in the PR checks section) and unblock them. If you do not have permission to unblock, ping simon-mo or khluu to add you in our Buildkite org.

Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging.

To run CI, PR reviewers can do one of these:

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

🚀