docs: clarify agent runtime demo scope

[hyeokjun32]

Summary

• Clarified that Lab's bundled Agent Runtime Reliability evidence is based on a synthetic/dummy sustained high-load 3-agent scenario.
• Kept real lightweight workload contention and device-specific sustained validation framed as next steps.

Validation

• poetry run python3 -m pytest tests/test_agent_runtime_report.py -q

Notes

• Documentation-only change.
• No Runtime result, compare, or deployment decision contract changed.