Small value sumcheck

[wu-s-john]

Note: This PR builds on the Barrett reduction infrastructure from #111. The changes span src/big_num/, src/lagrange_accumulator/, and src/small_sumcheck.rs. Reviewers are encouraged to focus on src/lagrange_accumulator/ and src/small_sumcheck.rs — the src/big_num/ changes are covered by #111.

Summary

Implement the small-value sumcheck optimization (Algorithm 6) from "Speeding Up Sum-Check Proving" (Bagad, Dao, Domb, Thaler, IACR 2025/1117). This replaces expensive field multiplications with native integer arithmetic during the first ℓ₀ rounds when polynomial values are guaranteed small.

This PR provides the core infrastructure for the optimization. Integration into the Spartan prover is planned for a follow-up PR.

Benchmarks

Measured on M1 Max MacBook Pro with jemalloc, BN254 scalar field, ℓ₀ = 3.

num_vars	n	baseline (DMR)	small-value	speedup
16	65,536	4.98 ms	3.43 ms	1.45×
17	131,072	8.83 ms	4.79 ms	1.84×
18	262,144	15.07 ms	7.54 ms	2.00×
19	524,288	27.81 ms	12.84 ms	2.17×
20	1,048,576	49.96 ms	26.37 ms	1.90×
21	2,097,152	190.66 ms*	50.73 ms	3.76×*
22	4,194,304	152.75 ms	96.09 ms	1.59×
23	8,388,608	285.86 ms	181.02 ms	1.58×
24	16,777,216	546.06 ms	333.13 ms	1.64×
25	33,554,432	1,095.2 ms	653.01 ms	1.68×
26	67,108,864	2,150.1 ms	1,314.5 ms	1.64×

These results are consistent with the ~1.6-2× speedup and has achieved better overall performance than #98.

Key Components

Lagrange accumulator infrastructure (src/lagrange_accumulator/):

• LagrangeAccumulators: precomputed A_i(v, u) values for all rounds
• LagrangeIndex/LagrangePoint/LagrangeHatPoint: type-safe domain indices
• LagrangeBasisFactory: barycentric Lagrange basis with O(D) evaluation
• extend_to_lagrange_domain: batch extension from {0,1}^ℓ₀ to U_D^ℓ₀
• EqRoundFactor: tracks α = eq(τ_{<i}, r_{<i}) across rounds

Performance optimizations:

• Thread-local SpartanThreadState eliminates per-iteration allocations
• Delayed modular reduction via SignedWideLimbs accumulators
• Skip binary betas (Az·Bz = Cz on {0,1}^n for satisfying witnesses)
• Batched eq-weighted binding in transition phase

API:

• SmallValue trait: WideMul + Copy + Zero + Add + Sub + Send + Sync
• SmallValueEngine<SV>: blanket impl consolidates field requirements
• prove_cubic_small_value<E, SV, const LB>: main entry point

Test Plan

• prove_cubic_small_value produces identical proofs to prove_cubic_with_three_inputs (equivalence tests)
• cargo test -- --skip test_msm_ux
• cargo clippy