More density stuff

[cscherrer]

MeasureTheory has these tests:

@testset "Density measures and Radon-Nikodym" begin
    x = randn()
    let d = ∫(𝒹(Cauchy(), Normal()), Normal())
        @test logdensity_rel(d, Cauchy(), x) ≈ 0 atol = 1e-12
    end

    let f = 𝒹(∫(x -> x^2, Normal()), Normal())
        @test f(x) ≈ x^2
    end

    let d = ∫exp(log𝒹(Cauchy(), Normal()), Normal())
        @test logdensity_rel(d, Cauchy(), x) ≈ 0 atol=1e-12
    end

    let f = log𝒹(∫exp(x -> x^2, Normal()), Normal())
        @test f(x) ≈ x^2
    end
end

The ones with let f pass just fine. But the let d ones fail. It seems like it's still kind of messy. What's the best way to get this working?