Function minimize 

pub fn minimize<F>(
    objective: F,
    initial_params: &Array1<f64>,
    config: &OptimizationConfig,
) -> Result<OptimizationResult, String>
where
    F: Fn(&Array1<f64>) -> Result<f64, String>,

Gradient-free coordinate-descent minimiser.

Performs real optimisation: at each iteration it tries a positive and a negative step along every coordinate, accepting any move that lowers the objective, and shrinks the step size when a full sweep makes no progress. It terminates on max_iterations, when the step size underflows tolerance, or when an entire sweep yields no improvement. This genuinely reduces the cost rather than echoing the initial parameters.