Remove filterpy stats whereever possible

[rlabbe]

For example, I wrote the function G=gaussian, and norm_cdf. Are these really necessrary, given the frozen rvs supported by SciPy. Even if SciPy is slightly less useful for my purposes, isn't it better for people to learn SciPy?

Another function that probably shouldn't exist is rand_student_t. I've implemented this both in chapter 3 and filterpy. scipy surely implements this.