Add rigid-body stability derivatives and a trim-angle search, as in the Python package #330

The Python package has rigid-body stability derivatives and a trim-angle search; this package has neither. The closest thing here is linearize (src/solver.jl:1266), which returns the dimensional Jacobian of forces and moments with respect to va and omega (about the origin), for one wing only.

What Python has, in src/VSM/stability_derivatives.py and src/VSM/trim_angle.py (awegroup/Vortex-Step-Method@a397bcd, v2.0.0):

• compute_rigid_body_stability_derivatives(body_aero, solver, angle_of_attack, side_slip, velocity_magnitude, roll_rate, pitch_rate, yaw_rate, step_sizes, reference_point, nondimensionalize_rates=True): dC{x,y,z}_d{alpha,beta,p,q,r} and dCM{x,y,z}_d{...} by central differences, rate derivatives optionally in p̂ = pb/2V, q̂ = qc/2V, r̂ = rb/2V.
• map_derivatives_to_aircraft_frame: VSM frame (x back, z up) to aircraft frame (x forward, z down).
• build_malz_coeff_table_from_solver: the derivatives as quadratics in alpha, the table format of Malz et al.'s kite models.
• compute_trim_angle(body_aero, solver, side_slip, velocity_magnitude, ..., alpha_min, alpha_max, reference_point): coarse alpha sweep for a CMy sign change, bisection, and dCMy/dalpha < 0 as the stability check.

Proposal: build both on linearize rather than on finite differences — the alpha/beta derivatives are the chain rule through va, and p/q/r already come out of the omega columns once the rotation is taken about the reference point (#329). Needs support for more than one wing in linearize to cover multi-wing bodies.

Upstream caveats worth not porting: build_malz_coeff_table_from_solver on Python main still passes the removed yaw_rate/pitch_rate/roll_rate keywords to va_initialize, and _evaluate_with_angles drops body_axis (fixed only on the improve_plotting branch).

Found while surveying the Python package for #113.