Added Binomial distribution

[MoritzNeuberger]

Source code is inspired by the scipy implementation.

[HDembinski]

Very nice contribution at first glance, sorry I missed this. I don't get notified by Github when a PR is created, but I get a notification if you ping me explicitly with @HDembinski

[HDembinski]

I think you want the type of p here, not n

[HDembinski]

trial

[HDembinski]

There were a couple of other issues, which I fixed here, that had quite far-reaching ripples into the code.

• The binom functions should accept both k and n arrays, while in your implementation n was a scalar. The idea that only the first argument is an array is deeply rooted in the design of the library, though, so it is natural to design it like this. To support the more flexible case, I revised the _jit utility to support functions with more than one array argument.
• You cannot compute the cdf by summing up the pmf, because that does not scale to large n. The correct way is to use the incomplete beta function.

[MoritzNeuberger]

Oh wow, you are right, that's pretty cool!