Output image incorrect when OffsetArray vectors passed as kernel components

If a 2d kernel is passed as two OffsetArray vectors, the second one transposed or adjoint, the final image is incorrect.

The issue does not occur if the second component is an OffsetArray matrix with one row, or if the vector is transposed /adjointed before being passed to centered.

image = [0.0 0.0 0.0; 0.0 1.0 0.0; 0.0 0.0 0.0]
image_out_typical = imfilter(image, Kernel.gaussian(1))

factor1, factor2 = ImageFiltering.factorkernel(Kernel.gaussian(1))
@info factor1
@info factor2
	
image_out_factors = imfilter(image, (factor1, factor2))

factor1a = factor1 |> vec |> centered
factor2a = (factor2 |> vec |> centered)'

@info factor1a
@info factor2a

image_out_factors_vec = imfilter(image, (factor1a, factor2a))
	
plot(heatmap(image, aspectratio = 1), heatmap(image_out_typical, aspectratio = 1), heatmap(image_out_factors, aspectratio = 1), heatmap(image_out_factors_vec, aspectratio = 1))

Start by running the provided imfilter example with Kernel.gaussian(1) and compare the typical, factor, and vector-factor outputs. Trace the handling of the two OffsetArray vector components, then add coverage showing that a transposed or adjointed second vector produces the same image as the equivalent one-row matrix and centered factor.