Improve `iterative_methods` structure

[rparini]

1. There is a certain amount of repetition in the stopping criteria for newton, secant and muller, these could probably be centralised to reduce repetition, perhaps in some stop_iteration function.
2. Would be nice to use some more battle-tested implementaions for these standard methods, rather than my own.
   • I think I originally passed on scipy.optimize.newton before because of https://github.com/scipy/scipy/issues/10103 but it has now been pointed out that casting x0 to np.complex128 works around that. I also make use of callbacks during the iteration in my cuurent implemention but could probably get the same effect by wrapping the target func passed to scipy.optimize.newton.
   • Or maybe threre`s some other capability in mpmath or numdifftools that could be used more easily.