solutions-engineering: compare fraud benchmark runner sizes

[adityamaru]

Summary

• Run the fraud benchmark as a 4-vCPU and 16-vCPU matrix.
• Use a fixed worker count so the larger runner shows CPU headroom instead of simply consuming all available CPU.
• Preserve the same fixed workload so candidates can compare duration and CPU metrics side by side.

Test plan

• Ran npm run lint in solutions-engineering/migration-lab/app.
• Ran PROFILE_CPU_TASKS=4 PROFILE_CPU_ITERATIONS=500000 PROFILE_CPU_WORKERS=2 npm run profile:cpu.
• Validated in the fork playground: 4-vCPU ran at CPU p95 100%, while 16-vCPU ran around CPU p95 41% with lower wall time.

Made with Cursor

---

^{Need help on this PR? Tag @codesmith with what you need.}

• Let Codesmith autofix CI failures and bot reviews (Staging)

[adityamaru]

Closing this version because the 16-vCPU comparison reveals the intended fix in the candidate workflow. Replacing with a smaller cleanup that keeps the undersized fraud benchmark as the problem signal.