Fix var/std along a dimension with UnitWeights

[devmotion]

var(A::AbstractArray, w::UnitWeights, dim::Int) — and therefore std along a dimension, which dispatches to it — compared the weight length against the total number of array elements (length(v)) instead of the size along the reduction dimension (size(v, dim)):

function var(v::AbstractArray{<:Real}, w::UnitWeights, dim::Int; mean=nothing,
                  corrected::Union{Bool, Nothing}=nothing)
    length(w) == length(v) || throw(DimensionMismatch("Inconsistent array lengths."))
    ...
end

As a result the method threw DimensionMismatch("Inconsistent array lengths.") for essentially any multidimensional array, even when the weight vector correctly matched the reduced dimension:

julia> using StatsBase

julia> var(reshape(1.0:12.0, 3, 4), uweights(3), 1)
ERROR: DimensionMismatch: Inconsistent array lengths.

This compares against size(v, dim) instead, consistent with the existing mean(A, w::UnitWeights; dims) method which already checks size(A, dims) == length(w).

weight_funcs in test/moments.jl is (weights, aweights, fweights, pweights) and does not include uweights, so the matrix var(x, w, dim) / std(x, w, dim) tests never exercised UnitWeights — hence the gap. Added a regression testset covering weighted var/std along a dimension with UnitWeights, including the mismatched-length error case.

Also bumped the patch version to 0.34.12.

🤖 Generated with Claude Code