Add fixed-domain barycentric interpolation for repeated-query univariate evaluation

Summary

Add a fixed-domain barycentric interpolation surface to CompPoly.CPolynomial.CLagrange for repeated-query univariate evaluation over a field.

For pairwise distinct nodes x : Fin n → R, store the barycentric weights
w_i = ∏_{j ≠ i} (x_i - x_j)⁻¹
once and evaluate by the classical second barycentric formula:

• if z = x_i, return y_i;
• otherwise return
  (∑ i, w_i * y_i * (z - x_i)⁻¹) / (∑ i, w_i * (z - x_i)⁻¹).

Correctness target

Prove that for every y : Fin n → R and z : R,
BarycentricDomain.eval y z = (Lagrange.interpolate Finset.univ x y).eval z.

Also derive the CompPoly-facing corollary
BarycentricDomain.eval y z = (CLagrange.interpolate Finset.univ x y).eval z,
together with the roots-of-unity specialization agreeing with CLagrange.interpolatePow.

Motivation

This supports the Phase 2 evaluation/interpolation track by separating one-time nodal preprocessing from per-query evaluation on a fixed domain.

Start with the CompPoly.CPolynomial.CLagrange surface and the proposed BarycentricDomain.eval definition. Read the existing Lagrange.interpolate and CLagrange.interpolatePow results, then establish the stated equality for all y and z and the roots-of-unity specialization. Done means the fixed-domain evaluation, CompPoly corollary, and specialization are proved.