feat(control): unify joint limits and bound trajectory dynamics

Use shared limits for all seven axes across MoveIt, URDF and FRI.
Apply Ruckig smoothing and continuous JTC spline checks, stretching
trajectory timing to respect velocity, acceleration and jerk limits.
Require fresh stationary feedback for direct trajectories and protect
preparation from concurrent submissions and stop requests.

Add regression tests, dependencies and bilingual FRI documentation.
Include the updated LIN speed-sweep scenario and current 60 ms filter
configuration. Acceleration and jerk limits remain project settings;
physical high-speed twitching is still under investigation.

src/iiwa_planning/test/test_prepare_trajectory.py

@@ -0,0 +1,29 @@
+import numpy as np
+import pytest
+from iiwa_planning.direct_trajectory import prepare_trajectory
+
+LIMITS = [[-2, 2, 1, 1, 2]]
+
+
+def test_dense_points_are_smooth_and_retimed():
+    t, q, v, a, scale = prepare_trajectory([0, .1, .2], [[0], [.5], [1]], LIMITS)
+    assert scale > 1
+    np.testing.assert_allclose(q, [[0], [.5], [1]])
+    assert abs(v[1][0]) > 0  # do not stop at every CSV waypoint
+    np.testing.assert_allclose([v[0], v[-1], a[0], a[-1]], 0, atol=1e-9)
+    assert t[-1] > .2
+
+
+@pytest.mark.parametrize('times,positions', [([0, 0], [[0], [1]]), ([0, 1], [[0], [float('nan')]]), ([0, 1], [[0], [3]])])
+def test_bad_trajectory_is_rejected(times, positions):
+    with pytest.raises((ValueError, RuntimeError)):
+        prepare_trajectory(times, positions, LIMITS)
+
+
+def test_time_scaling_preserves_quintic_between_waypoints():
+    from scipy.interpolate import BPoly
+    times, positions, velocities, accelerations, scale = prepare_trajectory([0,.1,.2], [[0],[.5],[1]], LIMITS)
+    curve = BPoly.from_derivatives(times, [[q,v,a] for q,v,a in zip(positions,velocities,accelerations)])
+    grid = np.linspace(times[0],times[-1],10001)
+    for order,cap in [(1,1),(2,1),(3,2)]:
+        assert np.max(np.abs(curve(grid,nu=order))) <= cap * 1.000001