Allow either DataValues or Missings for representing missing data

@davidanthoff I wanted to run this by you before preparing a PR.

I think it should be possible to represent missing data values in the result using either the DataValues package or the Missings package. Suppose that all the read_* functions had an optional Bool argument to control this choice with its value being stored in the ReadStatDataFrame. The handle_variable! function then ends with

    push!(ds.data, ds.useMissings ? fill(Union{jtype, Missing}(missing), ds.rows) : DataValueVector{ds.rows))

I think the only additional changes would be to use

dest::Union{DataValueVector{T}, Vector{Union{Missing,T}}}

in the signature of the readfield! methods and to add the Missings package.

Shall I prepare a PR for your consideration?

Trace the read_* functions, ReadStatDataFrame, handle_variable!, and the readfield! methods to understand how missing values are currently represented. Check how the Missings package would fit alongside DataValues and update the affected signatures and dependency. Done means callers can choose either representation and both forms work through the read path.