Question on `ScaledKernel`

[simsurace]

What is the advantage/use case of using a vector for the variance here?

KernelFunctions.jl/src/kernels/scaledkernel.jl

Lines 14 to 22 in 1831cc6

	struct ScaledKernel{Tk<:Kernel,Tσ²<:Real} <: Kernel
	kernel::Tk
	σ²::Vector{Tσ²}
	end
	function ScaledKernel(kernel::Tk, σ²::Tσ²=1.0) where {Tk<:Kernel,Tσ²<:Real}
	@check_args(ScaledKernel, σ², σ² > zero(Tσ²), "σ² > 0")
	return ScaledKernel{Tk,Tσ²}(kernel, [σ²])
	end

It seems to me that the allocations just hurt performance without any benefit.
Shall we get rid of it?

[theogf]

There was this PR #397 that aimed at doing it but it was never finished.