Modify the Python function from the previous assignment to use the relative truncation error as stopping criterion, i.e. using $\left|\frac{b_n-a_n}{b_n+a_n}\right|<\epsilon$.

Verify that you now need more iterations with tolerance $\epsilon=0.001$, but that you get an answer that is correct in at least 4 significant digits (instead of 3 significant digits with the usual stopping criterion of absolute truncation error).