WeightedCovMatrix with missing values fails.

[gdkrmr]

x = convert(Matrix{Union{Float64, Missing}}, randn(300, 10))
x[1, 1] = missing

fit!(WeightedCovMatrix(Float64), x, repeat([1.0], 300))

fails because eachrow returns a view, which makes _fit!(WeightedCovMatrix, ...) dispatch to the wrong methods.