Add load pattern configuration guide to benchmarks

[mpashkovskii]

Purpose

Add comprehensive documentation for load pattern configuration parameters in vLLM benchmarking tools. This PR enhances the benchmarks documentation by:

• Adding detailed explanations of --burstiness, --request-rate, and --max-concurrency parameters
• Providing mathematical foundations for traffic pattern generation using Gamma distributions
• Including data-driven visualizations showing request arrival patterns and inter-arrival time distributions
• Offering practical use case recommendations for different testing scenarios
• Explaining how to interpret KV cache configuration logs for capacity planning

This addresses the need for better guidance on simulating realistic load patterns for performance testing and capacity planning.

Test Plan

N/A

Test Result

N/A

---

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.

BEFORE SUBMITTING, PLEASE READ https://docs.vllm.ai/en/latest/contributing (anything written below this line will be removed by GitHub Actions)