feat(poi): improve benchmark latency measurement and reliability

[sainijit]

• Use VLM metrics files as primary latency source (matching SAD pattern)
• Prefer mqtt_received_at over frame timestamp in alerts API fallback
• Add early exit on first alert for single benchmark mode
• Add benchmark_duration parameter for configurable wait time
• Filter stale alerts using iteration start timestamp
• Skip container recreation when services already running
• Add dual latency method: single run vs stream density
• Clean metrics files before each iteration

PR Checklist

• Added label to the Pull Request for easier discoverability and search
• Commit Message meets guidelines as indicated in the URL https://github.com/intel-retail/performance-tools/blob/main/CONTRIBUTING.md
• Every commit is a single defect fix and does not mix feature addition or changes
• Unit Tests have been added for new changes
• Updated Documentation as relevant to the changes
• All commented code has been removed
• If you've added a dependency, you've ensured license is compatible with repository license and clearly outlined the added dependency.
• PR change contains code related to security
• PR introduces changes that breaks compatibility with other modules (If YES, please provide details below)

What are you changing?

Issue this PR will close

close: #issue_number

Anything the reviewer should know when reviewing this PR?

Test Instructions if applicable

If the there are associated PRs in other repositories, please link them here (i.e. intel-retail/performance-tools )

[Copilot]

Pull request overview

This PR updates the POI stream-density benchmark script to improve latency measurement fidelity and benchmark reliability by prioritizing VLM metrics files, improving alerts-based fallback, and adding a dedicated single-benchmark mode with configurable duration and early-exit behavior.

Changes:

• Prefer vlm_application_metrics files as the primary latency source, with a stream-density-specific extractor (last N pairs) and alerts API as a fallback.
• Improve alerts API parsing by preferring mqtt_received_at, filtering stale alerts using an iteration start timestamp, and applying the same filtering when saving thumbnails.
• Add single-run benchmark enhancements: configurable BENCHMARK_DURATION, early exit on first alert, and skipping container recreation when key services are already running.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.