Differentiate better between types of populations (be explicit when constructing KDE-based uncertain values)

[kahaaga]

Currently, UncertainValue(v) where v is some vector will trigger kernel density estimation of the distribution of v.

It would be nice to be able to differentiate between a simple population and a weighted population. Assume population is a vector and wts is a vector of weights associated with the elements of population. We'd then have the syntax UncertainValue(population) construct a EquiprobablePopulation(or something like that) and UncertainValue(population, wts) construct a WeightedPopulation (or something like that).

This way, a user can seamlessly use the resampling methods from Bootstrap.jl and Jackknife to sample equiprobably populations, while the fallback resampling method for the other uncertain value types is BasicSampling (see common interface suggestion in #24)