Higher-order mixture distribution

[WardBrian]

Discussed in stan-dev/design-docs#53, but doesn't require any additional new syntax:

An M component mixture where each component uses the same distribution function could be written with a variadic function as follows:

vector[N] y;
simplex[M] p;
vector[M] mu;
vector[M] sigma;

// mixture with M components
y ~ mixture_?(p, normal, mu, sigma)

Name is up for debate

[avehtari]

One possibility for the name would be family_mixture(), where family refers to the fact that in this case all components belong to the same family of distributions (e.g. normal)