fix: use CPU-aware prover scheduler

[jotabulacios]

This PR fixes the CPU proving regression introduced by the table scheduler.

The GPU scheduler uses external driver threads to admit and overlap table work. On CPU, that caused table tasks to launch nested Rayon work from outside the Rayon pool, leading to oversubscription, cache contention, and memory-bandwidth contention.

The CPU path now:

• schedules table work through the existing Rayon pool;
• limits table concurrency in bounded chunks; and
• separates auxiliary-trace construction/commitment from rounds 2–4 with a CPU phase barrier.

The CUDA path and proof format are unchanged.

[jotabulacios]

/bench

[github-actions]

Benchmark — real block (ethrex_mainnet_25368371.bin) (median of 3)

_{continuations · epoch 2^22 · 11 epochs}

Metric	main	PR	Δ
Peak heap	47250 MB	52967 MB	+5717 MB (+12.1%) 🔴
Prove time	141.389s	124.107s	-17.282s (-12.2%) 🟢

🎉 Improvement on the real block — prove time down 12.2%.

_{Prove-time spread 0.7% (123.672s / 124.107s / 124.480s)}

Memory Growth

_{ethrex distinct-account transfers · default parallelism · 1 sample per point}

Transfers	main (MB)	PR (MB)	Δ
4	14086	15070	+984 MB (+7.0%)
8	16094	19692	+3598 MB (+22.4%)
12	19602	24761	+5159 MB (+26.3%)
16	24994	30187	+5193 MB (+20.8%)
20	26996	32290	+5294 MB (+19.6%)

Growth rate: 1123 MB / transfer (main: 868, Δ: +29.4%)
Fit: R² = 0.9845 (main: 0.9749)

⚠️ Memory scaling regression — growth rate increased by +29.4%

_{Commit: b5256ee · Baseline: cached · Runner: self-hosted bench}

[diegokingston]

/bench

[MauroToscano]

/bench-growth

[MauroToscano]

Feedback incoming