document keywords (maxIter, x_tol, g_tol, …) better?

[stevengj]

Right now the documentation doesn't really seem to cover the keyword arguments, and never discusses cases that converge slowly (in which you might need to decrease the tolerances).

e.g. one such example is discussed on discourse here: https://discourse.julialang.org/t/alternatives-to-lsqfit-for-nonlinear-curve-fitting/107685/9?u=stevengj

[pkofod]

Thanks for bringing this to my attention.