Improved rational approximation algorithm that overrides the to_r method of Float
>= 0
Gregory H. Halverson