Function levenberg_marquardt_with_mask 

pub fn levenberg_marquardt_with_mask(
    model: &dyn FitModel,
    y_obs: &[f64],
    sigma: &[f64],
    params: &mut ParameterSet,
    config: &LmConfig,
    active_mask: Option<&[bool]>,
) -> Result<LmResult, FittingError>

Run the Levenberg-Marquardt optimizer with an optional per-bin active mask (SAMMY EMIN/EMAX-equivalent fit-energy-range restriction).

When active_mask is Some(m):

• bins where m[i] is false are excluded from χ² and the normal equations (weight zeroed before assembly), so they contribute nothing to the gradient or Hessian;
• the model is still evaluated on the full grid so resolution broadening at the boundaries of the active region is correct;
• dof = n_active − n_free where n_active is the count of active bins.

When active_mask is None, behaviour is identical to levenberg_marquardt (all bins active).