Fix ConstraintDual transformation for SplitHyperRectangleBridge

[blegat]

Fix for the issues detected in the SolverTests run of #2808
I did not add any test since the test is in #2810
I did #2810 in a separate PR not only to show in CI that it would have caught this one but also to avoid delaying this PR and hence #2808 in case #2810 requires some discussion

[odow]

I don't like that all the solver-tests are now going through this bridge. Is that what we want? Or do we need to change the bridge weight?

[odow]

https://github.com/jump-dev/MathOptInterface.jl/actions/runs/16919956500

[odow]

It looks like this fixed some errors, but there are a couple of other issues

[blegat]

I don't think it's an issue that this bridge is now used. If you want to go from Interval to Nonnegatives, it makes sense

[blegat]

If you go Interval -> LessThan and GreaterThan then you still need 2 bridges to create Nonnegatives. If you go Interval to Hyperrectangle, only one bridge is left. What do you think it's an issue that we use this bridge ?

[odow]

It makes sense that this bridge is chosen. It's just a large change (evidenced by the pretty broad failure of solver tests). But maybe that's okay. It's good if Hyperrectange gets more use