Function mpc_control 

pub fn mpc_control(
    a: f64,
    b: f64,
    state: f64,
    setpoint: f64,
    horizon: u32,
    u_min: f64,
    u_max: f64,
    steps: u32,
    control_penalty: f64,
) -> f64

One-step Model Predictive Control for a scalar LTI plant x_{k+1} = a·x_k + b·u: search candidate controls across [u_min, u_max] (a grid of steps+1 points), simulate horizon steps holding u, and return the u minimising Σ (setpoint − x_k)² + control_penalty·u². The optimal FIRST move of the receding horizon. Zero-heap (bounded loops, no allocation).