Bridge Newton hydroelastic contacts into kSAP

[Langford-H]

Summary

• adapt the example and runtime wrappers to Newton 1.5 target-array and CollisionPipeline APIs
• forward Newton per-contact Hydroelastic stiffness, damping, and friction into every kSAP contact-scatter path
• preserve Newton-generated Hydroelastic contact points while retaining rigid-contact fallback behavior
• report actual active/truncated contact counts and real per-world solver iteration and convergence summaries

Why

The existing Newton adapter carried contact containers but rejected Hydroelastic integration. Newton already generates pressure-derived distributed contacts; this change allows original kSAP to consume their effective material data without copying the solver or replacing Newton's owning pipeline.

Validation

• Python compilation passed
• Ruff E9,F63,F7,F82 checks passed
• full CUDA qualification on an RTX 4090 passed free flight, one-point contact, all eight matched Painleve cases, static Hydroelastic force/moment profiles, and one-step reduced/unreduced Hydroelastic-through-kSAP solves
• deepest unreduced case solved 1,396 contacts with zero truncation
• complete harness, raw evidence, provenance, and verdict: DISCO MR !2

Boundaries

This is an exploratory Newton 1.5 bridge, not a production compatibility claim. The retained qualification used Warp 1.17.0.dev20260727 while Newton locks Warp 1.16, and the Hydroelastic gate used zero damping. Dynamic Hydroelastic rollout, kSimilar, kLagged, performance claims, and production hardening remain outside this PR.