(AngleBasedLSCM) Improve construction of A and B matrices

[csparker247]

Summary

The A and B matrix assembly in AngleBasedLSCM::Compute is a large block of nearly-identical per-vertex conditionals repeated three times per face (for e0, e1, e2). This is hard to read and maintain.

Proposed Fix

Extract a helper lambda inside Compute:

auto addContrib = [&](std::size_t row, const EdgePtr& e, T c, T s) {
    if (e->vertex == p0)      { /* push to tripletsB (fixed p0) */ }
    else if (e->vertex == p1) { /* push to tripletsB (fixed p1) */ }
    else {
        auto freeIdx = freeIdxTable.at(e->vertex->idx);
        /* push to tripletsA (free) */
    }
};

Replace the three repeated per-vertex blocks with calls to addContrib. Add comments linking the structure to the equations in Lévy et al. (2002).

Acceptance Criteria

• No behavioral change — all existing tests pass with bit-identical results
• Face loop body is visibly shorter

Conductor Track

A6

---

Migrated from GitLab issue #2