Nudge the Line Into Place

A bakery tracks how many trays of bread it bakes (xs) and how many loaves it sells (ys). You want the straight line $\hat y = w x + b$ that fits them, and you will find it the way the lesson did: by walking downhill.

Write fit_line_by_descent(xs, ys, rate, steps).

• Start at w = 0 and b = 0.
• Each step, work out the errors w * x + b - y for every point, using the current w and b.
• From those same errors, work out both gradients: grad_w is the mean of 2 * x * error, and grad_b is the mean of 2 * error.
• Then move both knobs downhill: w = w - rate * grad_w and b = b - rate * grad_b.
• Repeat this exactly steps times. If steps is 0, nothing moves.

Return [w, b], each rounded to 3 decimal places, as plain Python floats.