fix: explicitly select CUDAExecutionProvider to avoid silent CPU fallback when TensorRT is absent (closes #5860)

[botbikamordehai2-sketch]

What

When force_cpu=False and a CUDA GPU is available but TensorRT is not installed, ONNX Runtime silently falls back to CPUExecutionProvider instead of using CUDAExecutionProvider. This happens because ORT's default provider priority list places TensorrtExecutionProvider before CUDAExecutionProvider, and when TRT fails to load it skips CUDA entirely.

Fix

Explicitly build the providers list in new_inference_session() by checking onnxruntime.get_available_providers() and preferring CUDAExecutionProvider when it is available and force_cpu=False. This ensures CUDA is used whenever possible, regardless of whether TensorRT is installed.

Closes #5860

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.

---

botbikamordehai2-sketch seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no potential bugs to report.

View in Devin Review to see 1 additional finding.