Add user-defined pre-modification of unknowns as well as update and check functions in Newton-Rhapson iterations

[adtzlr]

This should involve

• preprocess the unknowns y = pre(x) and by consequence, calculate f as f(y, *args); this defaults to pre = lambda x: x
• update the solution x = update(x, dx); this defaults to lambda x, dx: x + dx
• check if success success = check(dx, x, f); this defaults to check = lambda dx, x, f, tol=np.finfo(float).eps: np.linalg.norm(dx) < tol

With this modifications it should be possible to combine contique and felupe. The resulting newtonrhapson function should be very abstract and hence, the code should have good readability.