Fix/remove self loops adjmat mutation

[Parvm1102]

closes #658

Now the graph will be copied so that it does not mutate the original graph in place.

I have also added tests for remove_self_loops for dense graph as well

[Parvm1102]

@CarloLucibello , the GNNLux checks are failing, they were also failing in the last few open prs.
I think it is related to some recent upstream changes thats triggering Automooncake() tests.

[Parvm1102]

@CarloLucibello , the GNNLux checks are failing, they were also failing in the last few open prs.
I think it is related to some recent upstream changes thats triggering Automooncake() tests.

in #626 ,

there was a compat added for GNNLux:

[compat]
LuxTestUtils = "1"

However it was removed in a later pr, we may enable it again or we may also explicitly skip AutoMooncake() here:

GraphNeuralNetworks.jl/GNNLux/test/test_module.jl

Line 74 in b9cdc0f

test_gradients(loss, x, ps; atol, rtol, skip_backends=[AutoForwardDiff(), AutoEnzyme()])

[CarloLucibello]

we may enable it again or we may also explicitly skip AutoMooncake() here:

yes we could skip Mooncake for those tests that are failing at the moment. thanks for looking into it.