implement midpoint derivative calculation

[Swanson-Hysell]

Currently within ipmag.curie() there is code to calculate the derivative which is then smoothed. It seems preferable to use the midpoint temperature (i.e. T_mid = (T[i + 1] + T[i]) / 2) for the derivative rather than the original T array values in order to associate the derivative value between the two points from which it is calculated.

Here is the current relevant code:

    # calculate first derivative
    d1, T_d1 = [], []
    for i in range(len(M_smooth) - 1):
        Dy = M_smooth[i - 1] - M_smooth[i + 1]
        Dx = T[i - 1] - T[i + 1]
        d1.append(Dy/Dx)
    T_d1 = T[1:len(T - 1)]
    d1 = np.array(d1, 'f')
    d1_smooth = smooth(d1, window_len)