pdf vs. pmf

[OAGr]

Right now we support pdf() on distributions. One problem is that discrete distributions instead use/require pmf; probability mass functions.

We could support two different functions, pdf() and pmf(). We might also want to support some third function, which returns both; like,

probabilityAt(mx(3,normal(1,5)), 3) -> {mass: 0.5, density: 0.2}

My guess is that this will be a bit confusing to users.