simulation: derive the stepper latch from moving bodies on add/remove

One helper sets stepper.blocked from _moving_bodies() after add_body,
pre-init remove_body, initialize and stepper recreation, so the latch never
disagrees with motion ownership. Tests: latch follows add/remove; the
after-close assertion checks the actual refusal.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

tests/unit/test_moving_bodies.py

@@ -232,6 +232,16 @@ def test_add_body_and_set_body_route_targets(monkeypatch):
                          **kwargs)
 
 
+def test_stepper_latch_follows_moving_bodies(monkeypatch):
+    sim = routed_simulation(monkeypatch)
+    still = sim.add_body("circle", (100., 40.), 5.)
+    assert sim.stepper.blocked is None
+    moving = sim.add_body("circle", (50., 40.), 10., translation=Translation((0., 0., -1., 1.), .1))
+    assert sim.stepper.blocked is not None
+    sim.remove_body(moving)  # removing the last translating circle clears the latch
+    assert sim.stepper.blocked is None and sim.bodies.get(still).motion is None
+
+
 def validate(monkeypatch, objects, flags=None, **cfg):
     sim = routed_simulation(monkeypatch)
     for obj in objects:
@@ -306,5 +316,5 @@ def test_failed_moving_step_marks_simulation_failed(monkeypatch):
     with pytest.raises(RuntimeError, match="unknown"):  # no retry on the failed circle
         motion.advance(lambda *_: None, lambda: None, lambda: None)
     sim.close()  # close still works
-    with pytest.raises(Exception):  # no re-entry after close
+    with pytest.raises(RuntimeError, match="Simulation failed"):  # no re-entry after close
         sim.run(1)