Function coherence_score 

pub fn coherence_score(matrix: &dyn Matrix) -> Precision

Compute the diagonal-dominance margin of a sparse matrix.

For each row i, computes |diag[i]| - Σ_{j≠i} |A[i,j]| (the “diagonal-dominance excess”) and divides by |diag[i]| to get a dimensionless score. The matrix’s coherence is the minimum of these per-row scores: the worst row dominates the bound.

Returns a value in [-∞, 1]:

• 1.0 — perfectly diagonal (every off-diagonal is zero).
• (0, 1) — strictly diagonally dominant; the larger the value, the more coherent. Neumann series convergence is guaranteed iff > 0.
• 0.0 — exactly on the diagonal-dominance boundary.
• negative — not diagonally dominant; iterative solvers may diverge.

Cost: one pass through the matrix’s row iterator. O(nnz(A)) — matches Linear complexity class per ADR-001.