Add prover benchmark to measure scaling

[gabrielbosio]

Adds benchmark script that runs different sizes of Fibonacci program and measures per-phase timing and heap peak and shows how it scales when increasing program size.

[gabrielbosio]

/bench

[github-actions]

Benchmark — fib_iterative_8M (median of 3)

_{Table parallelism: 32 (auto = cores / 3)}

Metric	main	PR	Δ
Peak heap	68340 MB	68016 MB	-324 MB (-0.5%) ⚪
Prove time	33.922s	33.866s	-0.056s (-0.2%) ⚪

✅ No significant change.

⚠️ Baseline heap spread: 5.8% (66969 MB / 68340 MB / 70933 MB) — comparison may be less reliable

_{Commit: 8828e65 · Baseline: cached · Runner: self-hosted bench}