Find a better name for `real` in `Continuous`

[avibryant]

As of 0.3.0rc2, the way to create a Real random variable from any Continuous distribution is with the real method.

As I work on multivariate distributions, this seems short-sighted, since a MVNormal distribution's random variable would be a Vec[Real], and real is not an awesome name for a method that returns that.

Options might include:

• keep real as the method for univariate distributions (with vec(k) for vectors of independent rvs), and use vec as the method for an MV(k) distribution.
• use param for both
• use rv or randomVariable for both

... something else?

[avibryant]

This obviously needs to be resolved before the final 0.3.0 release since it would be a breaking change after that.

[avibryant]

One argument against using rv is that generator is also producing a (different kind of) random variable from the same distribution.

An argument against using param is that the Real you get is not necessarily (and in fact almost never) the same as the raw, untransformed parameter that HMC will be working with.

[avibryant]

I'm currently really liking @snoble's suggest of latent