CategoricalArrays ("Factor") variables support for population proportions and counts

[smishr]

In R survey, population proportions and counts are calculated using mean and total respectively. Previously in Survey.jl we followed this approach, which can be seen in AnalyticalSolution branch.

I am wondering if a more "Julian" approach would be to have prop and count (fn name could be different) functions separately? This way we dont clutter mean and total estimators too much with extra if-else ladders or methods, and have clean work for prop and count specific functionalities.

Thoughts?

See #52 for previous discussion on categorical arrays

[smishr]

@ayushpatnaikgit what do you think?