Fix VehicleSim QSS initialization from FourPost state

[rhorvath02]

Problem

VehicleSim did not start from a coupled static suspension pose. The chassis reference frame used a legacy mass-weighted estimate while the four independent lower-arm coordinates started at zero. That combination did not reproduce the settled front/rear spring lengths or tire loads, so the simulation spent its opening transient finding the suspension equilibrium.

The old mass/CG calculation also represented only the supplied left-side axle records. This PR separates that legacy start-pose estimate from the actual complete-vehicle mass properties instead of treating the estimate as the vehicle CG.

QSS initialization

• Adds a vehicle-owned QSSInitializationRecord containing the reference velocity, chassis position and pitch, four lower-arm coordinates, solved spring lengths, and corner-load targets.
• Uses gravity-enabled FourPostSim suspension solutions at the settled front and rear corner loads to obtain the physical lower-arm coordinates and spring lengths.
• Initializes chassis heave and pitch consistently with those four contact-patch states and the compliant ground model.
• Fixes the selected lower-arm coordinates and initializes their angular velocities to zero in VehicleSim only. Other chassis users retain opt-in defaults.
• Exposes all four spring lengths in the standard simulation output.
• Calculates complete-vehicle mass, CG, and inertia from the sprung mass plus both axle sides.

The record is solved for the standard initialVel = 15 m/s; VehicleSim warns when used at another initial velocity.

Post-rebase VehicleSim results

Validated after rebasing onto current main, including the merged wheel alignment, force-frame, aerodynamic, transient-slip, vector-angle, and initialization-gate fixes.

Startup metric	Baseline	QSS initialization	Change
Initial total tire-load error	136.178 N	0.04396 N	99.97% lower
Peak absolute total-load error, first 2 s	249.081 N	17.746 N	92.88% lower
RMS total-load error, first 2 s	52.535 N	6.133 N	88.33% lower
Minimum tire load during maneuver	182.608 N	182.827 N	No material change

Initialized spring lengths:

Axle	FourPost target	VehicleSim at t = 0	Absolute error
Front, both sides	0.175156854189 m	0.175156852337 m	1.85 nm
Rear, both sides	0.244911388707 m	0.244911388690 m	0.017 nm

Initial loads remain left/right symmetric: 710.496 N per front corner and 721.413 N per rear corner. All recorded outputs are finite. Initialization completed without homotopy, and the standard ramp-steer simulation terminated normally at the QSS plateau at 6.140 s.

The corrected complete-vehicle mass properties remain:

• Mass: 261.07265114 kg
• CG: {-0.8002708499, 1.511554e-6, 0.2796156784} m

The post-rebase Windows run reported 37.31 s in the simulation stage. End-to-end time was dominated by a fresh OpenModelica backend translation, so runtime is not the primary claim; the material improvement is removal of the artificial suspension startup transient.

Validation

• Rebased conflict-free for all production Modelica files; the sole Python test conflict was resolved by preserving both the wheel-mirroring and QSS/mass tests.
• Ruff: pass.
• Fast Python suite: 31 passed; 108 OpenModelica-dependent cases skipped in that intentionally non-OMC pass.
• Combined QSS/mass and nonzero wheel-mirroring OpenModelica checks: 6 passed.
• Initialization audit: 33 unchanged fixtures passed; VehicleSimAnimationOn was rerun exactly and its one affected baseline row was refreshed for four new spring-length outputs and the QSS pose.
• Full standard VehicleSim ramp-steer run: pass; initialization without homotopy; finite outputs; normal QSS termination.

Out of scope

The remaining initial longitudinal acceleration is associated with drivetrain/controller torque balance. It is separate from the suspension heave/settling issue addressed here.