Least Squares Line Fit with Outlier Rejection

Fit a straight line to noisy measurements and iteratively discard outliers more than three standard deviations from the fit.

Constraints

• Function: [slope, offset, kept] = robust_fit(x, y).
• Iterate until no points are removed or 10 iterations pass.
• Reject points beyond 3 sigma of the residuals.
• Must not remove so many points that fewer than 2 remain.
• Return the indices of the kept points.
• Implement the fit directly, not with polyfit.