Adds sample default values for Binomial & Categorical random variables

[saresend]

What

This PR adds some default implementations for categorical and binomial distributions. Further, it also adds an owned initializer to the categorical random variable so that it is slightly more ergonomic to write code like the following:

let categorical = distribution::Categorical::new(vec![....]);

[IvanUkhov]

I am really uncomfortable adding that kind of defaults. Sorry.

But what is your use case? What do you need Default to be implemented?

[saresend]

My use case is that I'd like to wrap collections of random variables that model some degree of dependence on each other. While using certain distributions these collection types implement Default, but its a bit inconsistent in that changing some of the internal distributions can break that assumption. Of course I could wrap these things myself, but it felt a bit cleaner to let all distributions implement a default type that provides some non-trivial instance, so that it everything becomes derivable, rather than have to handle things manually.

[saresend]

But yeah, I understand your concern - I'm not sure if there'd be another trait that denotes less of a "standard" instance, and rather some instance that's convenient to construct?

[IvanUkhov]

My use case is that I'd like to wrap collections of random variables that model some degree of dependence on each other. While using certain distributions these collection types implement Default, but its a bit inconsistent in that changing some of the internal distributions can break that assumption.

Why is it necessary for your higher-level abstraction to have the underlying distribution implement Default? Perhaps it is imposed artificially and is not actually needed? The distributions are probably stored in some Vec, and Vec implements Default regardless of what is inside. Or you could implement Default conditionally:

impl Default for MyRandomVector<T>
where
    T: Distribution + Default,
{
}