fix(training): correct non-forward penalty math and add a grounding incentive

Adversarial review of the previous stage-4 retune found two problems:
non_forward_speed used planar_speed - forward_component, which under-charges
diagonal motion relative to true lateral speed (e.g. ~29% penalty at 45
degrees off the nose instead of the correct ~71%); fixed to the Pythagorean
magnitude for forward-facing angles, full speed for backward-facing ones.

Also, ground_tilt_penalty and non_forward_penalty only ever cost reward near
the floor with nothing offsetting them above it, which could teach a policy
that's still bad at ground handling to just avoid the floor rather than get
better at it. Added grounded_upright_reward (ship_ai_controller.gd) plus a
new ShipObservations.is_floor_contact helper for genuine belly-on-floor
contact detection, so grounding well while upright is the locally profitable
choice, not just the least-punished one.

Game/scripts/ship_observations.gd

@@ -143,3 +143,21 @@ static func contact_normal(ship: Ship) -> Vector3:
 		if normal.y < FLOOR_NORMAL_MIN_Y:
 			return normal
 	return Vector3.ZERO
+
+
+# True belly-on-floor contact — the complement of contact_normal, which
+# deliberately excludes floor contact (see its comment). Used by
+# ShipAIController.grounded_upright_reward to reward genuinely resting on
+# the floor rather than just being below the GROUND_HANDLING_HEIGHT proxy
+# altitude, so a ship can't collect ground-handling reward by hovering just
+# under the threshold without ever touching down.
+static func is_floor_contact(ship: Ship) -> bool:
+	var state := PhysicsServer3D.body_get_direct_state(ship.get_rid())
+	if state == null:
+		return false
+	for i in state.get_contact_count():
+		if not state.get_contact_collider_object(i) is ArenaBoundary:
+			continue
+		if state.get_contact_local_normal(i).y >= FLOOR_NORMAL_MIN_Y:
+			return true
+	return false