Fix failing yolo26 gpu integration tests

[mkaic]

What does this PR do?

When I implemented YOLO26 integration tests in inference_models, I failed to account for cases where the models produce different outputs depending on whether they are run on CPU or GPU. I have added branches dependent on DEFAULT_DEVICE to these tests with different expected values for CPU and GPU. Most cases are very slight differences in output confidence or bbox coordinates, but in a small number of cases, significant swings in prediction confidence are observed. I am not sure what the cause of this is, but I suspect it may have to do with different input preprocessing between branches. I have not been able to verify this though.

Type of Change

Fixes the GPU tests that I caused to fail accidentally in my YOLO26 PR.

• Bug fix (non-breaking change that fixes an issue)

Testing

I have run the GPU Integration Tests through Github actions and verified that they no longer fail on this branch.

• I have tested this change locally
• I have added/updated tests for this change

Test details:

See prior PR

Checklist

• My code follows the style guidelines of this project
• I have performed a self-review of my own code
• I have commented my code where necessary, particularly in hard-to-understand areas
• My changes generate no new warnings or errors
• I have updated the documentation accordingly (if applicable)

Additional Context

N/A