allow transform to avoid Z-score transforming when sigma=0

[tylerjthomas9]

Ref: #910

Here is my attempt to prevent NaNs when transforming data with 0 variance.

Here is the result:

julia> X = ones(5,8)
5×8 Matrix{Float64}:
 1.0  1.0  1.0  1.0  1.0  1.0  1.0  1.0
 1.0  1.0  1.0  1.0  1.0  1.0  1.0  1.0
 1.0  1.0  1.0  1.0  1.0  1.0  1.0  1.0
 1.0  1.0  1.0  1.0  1.0  1.0  1.0  1.0
 1.0  1.0  1.0  1.0  1.0  1.0  1.0  1.0

julia> feature_transform = fit(ZScoreTransform, X, dims=1)
ZScoreTransform{Float64, Vector{Float64}}(8, 1, [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0], [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0])

julia> StatsBase.transform(feature_transform, X)
5×8 Matrix{Float64}:
 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0
 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0
 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0
 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0
 0.0  0.0  0.0  0.0  0.0  0.0  0.0  0.0

[andreasnoack]

I don't like the discontinuity that this introduces. You'd end up with a huge difference between the transform of X and the smallest perturbation of `X.