[Enhancement]: rtol interface for tests

[ThijsJenneskens]

Feature/Enhancement Description

Linking #280: I would like relative tolerance as an option for @test_rules the same way absolute tolerance is, such that I can write:

@test_rules [check_type_promotion = true, rtol = 1e-4] ...

Motivation / Use Case

I'm writing some tests for extreme cases of the Gamma node, and the tests fails with cases like this:

┌ Warning: Testset for rule Gamma(:α, Marginalisation) has failed!
│ Inputs: (q_out = PointMass(x), q_θ = ReactiveMP.BayesBase.convert_paramfloattype(__promoted_T, Gamma(z, w)))
│ Expected output: ReactiveMP.GammaShapeLikelihood{Float64}(p=1.0, γ=-23.025850929890456)
│ Actual output: ReactiveMP.GammaShapeLikelihood{Float64}(p=1.0, γ=-23.02584966210056)

Proposed Solution

I would like relative tolerance as an option for @test_rules the same way absolute tolerance is, such that I can write:

@test_rules [check_type_promotion = true, rtol = 1e-4] ...

Alternatives Considered

No response

Example Use Cases

No response

Priority (from your perspective)

Important

Related Issues / Discussions

No response

Additional Context

No response

[Nimrais]

Yeah such interface would be nice.