automatic definition of how many beta components

[danielinteractive]

Design for:

• [ ] E.g.: start with 1 component, 2 components
• [ ] Is AIC better with 2? then also try 3
• [ ] otherwise stop
• [ ] https://pages.mtu.edu/~shanem/psy5220/daily/Day20/Mixture_of_regressions.html might be helpful
• [ ] it can also be helpful that we can pass multiple mixture distributions as starting points to fit_mle() , see https://github.com/dmetivie/ExpectationMaximization.jl/issues/9