eltype for Distribution type is ambiguous

[bvdmitri]

Now we use the Base.eltype option to determine the sample type of some Distribution. For example:

• eltype(NormalMeanVariance(...)) === Float64
• eltype(MvNormalMeanCovariance(...)) === Vector{Float64}
• eltype(Wishart(...)) === Matrix{Float64}

This leads to ambiguities with the SampleList "distribution", which defines eltype both over its sample type and weight type, e.g.:

• Tuple{Float64, Float64}
• Tuple{Vector{Float64}, Float64}
• Tuple{Matrix{Float64}, Float64}

We cannot really change the behaviour of the eltype function for the SampleList, because eltype is used by the Julia language itself for iteration.

To circumvent this ambiguity we probably need to implement sampletype function, where we can fallback:

sampletype(::Type{ T }) where {T} = eltype(T)

[bvdmitri]

Fixed in #229