[tomo0613]

Follow-up to #950: #980 merged the rolling-direction fix (#925) and the skid clamp (#946), but the two other behaviours from #950 are still missing.

• Free spin: a wheel with engine force doesn't spin up while airborne. It keeps coasting from its last ground speed, and the 0.99 damping even slows it down while the throttle is held. It also doesn't spin faster than the ground when it loses grip under power (burnout).
• Wheel lock: a braked wheel keeps rolling at ground speed while the vehicle skids, so it never visually locks.

#950 handled these with cannon-es' approach: a new free_spin_speed tuning parameter, and locking whenever brake > engine_force. That approach locked the wheel during any gentle braking. This PR derives both behaviours from the existing engine and brake impulses instead, with no new field or constant.

Approach

Each wheel's spin is driven by the engine and brake impulses acting on its share of the chassis mass (chassis.mass() / num_wheels). That is the same response the engine force has on the vehicle while the wheel has traction, so the spin stays continuous when the wheel leaves the ground, loses grip, or lands. The previous step's spin speed is recovered from the existing delta_rotation.

Wheel state	Spin (tire surface speed)
In contact, full traction (skid_info == 1)	ground speed (unchanged)
In contact, slipping, engine_force != 0	previous spin + (engine impulse − transmitted forward_impulse) / mass, never slower than the ground
In contact, skidding, brake != 0	previous spin slowed by (brake − transmitted forward_impulse) / mass toward zero, never faster than the ground
Airborne, engine_force != 0	previous spin + engine impulse / mass
Airborne, braking	previous spin slowed by brake / mass toward zero

As in update_friction, the brake only acts when engine_force == 0, and it is treated as a per-step impulse. The 0.99 free-spin damping now applies only when the engine isn't driving the wheel. Otherwise it caps the airborne spin at ~99× the per-step spin-up, so a wheel that leaves the ground at speed spins down under throttle.

Gentle braking (within the friction circle) keeps the wheel rolling with the ground. Only an actual skid locks it.

Tests

crates/rapier3d/tests/vehicle_wheel_free_spin.rs:

• airborne_wheels_spin_up_under_engine_force: spin keeps increasing under throttle, reverses with the engine force, and decays when the throttle is released.
• wheels_spin_faster_than_the_ground_on_burnout
• airborne_brake_stops_the_wheel_spin
• hard_braking_locks_the_wheels_while_sliding: wheels stop rotating while the chassis still slides.
• gentle_braking_keeps_the_wheels_rolling

The existing issue_925_wheel_rotation and issue_946_vehicle_braking_friction tests still pass. I also tested the change through the TypeScript bindings (a locally built rapier3d-compat) in a three.js driving game.