[Perf #478] LinearSVC: close liblinear-scale fit gap on Iris and Digits

Parent: #478

Current canonical v2 losses are real and large: Iris is ~0.613x end-to-end (Flow fit 1.580 ms vs sklearn 0.863 ms; prediction is already much faster), and Digits is ~0.216x (Flow fit 1650.7 ms vs sklearn 356.8 ms). The bottleneck is therefore overwhelmingly fit/solver work, not API overhead.

Audit and profile the LinearSVC training implementation against sklearn/liblinear. Attribute time to coordinate/gradient updates, multiclass strategy, convergence checks, memory layout, repeated dot products, allocation/copies and stopping criteria. Preserve canonical accuracy/parity.

Acceptance: dedicated fit flame/profile breakdown; operation-count comparison against sklearn/liblinear where practical; no avoidable per-sample/per-feature allocation; scaled 100/1k/10k+ sample benchmarks; Iris and Digits end-to-end >=0.95x first, target >=1.05x; prediction must not regress; learned weight/intercept state parity must be reported.

Start with the LinearSVC training implementation and profile fit separately on Iris and Digits. Compare coordinate or gradient updates, multiclass work, convergence checks, dot products, allocation and copies against sklearn/liblinear, then benchmark 100, 1k and 10k+ samples. Done means a dedicated profile, operation-count comparison, preserved accuracy and state parity, no avoidable per-sample or per-feature allocation, and no prediction regression.