Integrate redesigned Efficient Sumcheck

[z-tech]

What does this PR do?

• Integrates a large-scale redesign in Efficient Sumcheck: Major rewrite efficient-sumcheck#99

Based on microbench: Goldilocks hashchain, n=4096, l1=4, s=8, t=7, hash_chain_size=10, 30 iters + 2 warmup.

Phase	Original	Current	Change
rs_encode (FFT)	~2.5ms	~2.1ms	-16%
pesat_merkle_tree	~1.5ms	~1.5ms	—
twin_constraint_sumcheck	~7.2ms	~2.2ms	-69%
eval_bundled_r1cs	~0.6ms	~0.2ms	-67%
merkle_commit	~1.0ms	~1.0ms	—
eq_poly_evals + ood_evals_vec	~3.9ms	~0.2ms	-95%
batching_sumcheck (inner product)	~0.06ms	~0.06ms	—
Total (constrained_code_accumulate)	~12.9ms	~3.8ms	-71%
Total (incl. pesat_reduce)	~16.9ms	~7.4ms	-56%

I have a very large rewrite coming after this btw.

[z-tech]

Pr looks big but it's really only ~100 lines that are different and like a few hundred that are a tiny bit modified to use the API.

Bucket	Insertions	Notes
Import block rewrite	~20	pure line-for-line swap (old effsc paths → new)
TwinConstraintEvaluator struct+impl	~50	replaces the 40-line twin_constraint_round_poly closure in main — net **+10**
accumulate_sparse_evaluations + batched_constraint_poly	~35	genuinely new — effsc dropped them
prove() sumcheck call-site rewrites	~15	coefficient_sumcheck(|tw, pw| …) → CoefficientProverLSB::new + sumcheck(…); same for inner-product
prove() eq-evals helper swaps	~20	per-point loops → compute_hypercube_eq_evals(…)
verify() transcript reads	~3	two helper calls (pre-TC, between)
verify() sumcheck_verify + oracle checks	~25	replaces the per-round manual reductions in main (~20 deletions on the other side)
verify() misc (MSB→LSB reverse, proximity reorder, etc.)	~5	needed because InnerProductProver is MSB half-split
Doc comments and blank lines	~80	inflation from the above — every refactored hunk took more space in code-review-style comment blocks

I think the right decision in the end is to not have accumulate_sparse_evaluations + batched_constraint_poly in effsc but if it comes up again we could consider something like a crate for hyperplonk-utils.

[winderica]

Super sorry for the delay & awesome work as always!

The changes make perfect sense, and I only have a few (nitpicky) comments about removing unnecessary memory copies, which may or may not make things faster. It's totally fine to ignore them if you think the improvement is marginal!

[z-tech]

Thanks again for the very helpful comments @winderica ! Merging this now