solutions-engineering: simplify fraud sizing exercise

[adityamaru]

Summary

• Remove the idle background profile job so the sizing exercise focuses on the CPU-bound fraud benchmark.
• Keep the fraud benchmark on the initial 4-core GitHub runner label, but cap worker count so upsizing later can show CPU headroom.

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.

Made with Cursor

---

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

• Let Codesmith autofix CI failures and bot reviews

---

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

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