[raorjun]

Why

Tests/modelica_physics_baseline.csv checks only the final row. A FourPostSim closure residual grew linearly with time to 3.8 N at 118 s, and no test caught it until #48 fixed it. This test checks every output row, so a drift like that fails CI.

What changes

• Add Tests/test_four_post_closure.py. It simulates BobLib.Experiments.Standards.FourPostSim over its default 0 to 118 s heave and roll sequences.
• On every row, per axle, the world-vertical contact-patch forces plus <axle>KnC.jackingForce must sum to less than 1e-3 N. The test resolves <axle>AxleDW.<side>CP.f to world z through R.T. Tire Fz is clamped at zero during lift-off (t = 42 to 56 s), so Fz cannot show closure on those rows.
• On rows where all heave, roll, fx and fy commands are zero, <axle>ChassisActuator.sprungLoads.frame_a.f[1] and f[2] must be below 1e-3 N, and t[1] must be below 1e-3 N·m.
• The test_*.py wildcard puts it in make test-additional (the OpenModelica Component Contracts job). No makefile, CI or model changes.

Evidence

• The tolerances are about 1e-6 of the 1 kN static axle load. The measured floor on main is 2.7e-12 N or less for closure, 2.3e-13 N or less for f[1] and f[2], and 2.3e-10 N·m or less for t[1].
• On 2065e2b, the parent of the [Architecture] Integrate configurable axles and nested vehicle calibration #48 fix, the test fails: "fr vertical closure = 3.823e+00 at t = 118.0 s".
• On main 7363fc1 it passes in 171 s. Most of that is translation. The simulation takes about 2 s.
• make lint and make modelica-lint pass. Measured with OMC 1.27.1, MSL 4.1.0 and VI 2.0.2.

Limits

• It covers the default FourPostSim record and architecture (DWBCStabar) only.
• It adds about 3 minutes to the component job.
• The full make test-additional was not run locally. CI runs it.
• Finding along the way: at heave +22.9 to +38.1 mm the tires lift off, and the fixture pulls the wheels down with up to 1553 N of jacking force. Reported Fz is 0 there.