[Perf #478] Lasso fit: beat sklearn coordinate descent rather than settling for a near tie

Parent: #478

Canonical v2 Diabetes: Flow fit 0.784 ms vs sklearn 0.704 ms; Flow predict is much faster (0.0087 ms vs 0.0725 ms). End-to-end is ~0.980x, inside practical-near-tie territory but still below the #478 target.

Profile coordinate updates, residual recomputation, feature norms, convergence/duality-gap checks, memory access, zero-skipping and allocation. Audit whether residuals and norms are maintained incrementally and whether the feature loop vectorizes.

Acceptance: fit-only profile and iteration count comparison; scaled rows/features/sparsity/alpha sweeps; coefficient/intercept and objective-state parity; fit >=1.05x sklearn on the canonical case without regressing prediction.

Start with a fit-only profile of the Lasso coordinate updates, residual and feature-norm handling, convergence checks, memory access, zero-skipping, and allocation. Compare iteration counts on the canonical v2 Diabetes case, then run scaled rows/features/sparsity/alpha sweeps and check coefficient, intercept, and objective-state parity. Done means fit reaches at least 1.05x sklearn on the canonical case without regressing prediction.