Perf/cpu lde rework with default hasher

[MauroToscano]

Based on PR #650, with the PR-added deterministic FxHasher removed so op-dedup maps use Rust default keyed HashMap. This keeps the LDE and trace parallelization changes intact for benchmarking the hasher delta. If benchmark timing is similar, this avoids the deterministic-hasher DoS concern.

[MauroToscano]

/bench

[github-actions]

Benchmark — ethrex 20 transfers (median of 3)

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

Metric	main	PR	Δ
Peak heap	79872 MB	81977 MB	+2105 MB (+2.6%) ⚪
Prove time	50.360s	44.776s	-5.584s (-11.1%) 🟢

🎉 Improvement detected — heap or time decreased by more than 5%.

✅ Low variance (time: 1.8%, heap: 1.3%)

_{Commit: d1bf465 · Baseline: built from main · Runner: self-hosted bench}