feat(data): Default synthetic samples to max_requests

[git-jxj]

Summary

This PR improves the user experience of synthetic data generation by providing a more intelligent default for the number of samples.

The Problem:

Currently, when using synthetic data (e.g., --data='{"prompt_tokens": ...}'), the number of unique samples to generate defaults to 1000. If a user
only intends to run a small number of tests (e.g., --max-requests=50), the tool still spends a very long time generating 1000 large samples, only
to use 50 of them. This leads to a long, unexpected delay at the "Creating request loader..." step.

The Solution:

This PR plumbs the max_requests value down to the SyntheticDatasetCreator. The logic is updated so that if samples is not explicitly defined in
the --data configuration, it will default to the value of max_requests.

This makes the behavior much more intuitive:

• If a user runs --max-requests=50, the loader creates 50 samples.
• If a user runs --max-requests=50 but specifies --data='...samples=5', the loader respects the user's choice and creates only 5 samples.
• If max_requests is not set, the loader falls back to the original default of 1000.

This change significantly reduces the startup time for benchmarks that use synthetic data with a specified number of requests.

Details

• [ ]

Test Plan

guidellm benchmark run --max-requests 50 --data='{"prompt_tokens":100, "output_tokens":100}'

Before this PR: Observe the long delay and the log message: Created loader with 1000 unique requests...
After this PR: Created loader with 50 unique requests...

Related Issues

#319

• Resolves #

---

• "I certify that all code in this PR is my own, except as noted below."

Use of AI

• Includes AI-assisted code completion
• Includes code generated by an AI application
• Includes AI-generated tests (NOTE: AI written tests should have a docstring that includes ## WRITTEN BY AI ##)