Import onnxruntime before loading the TensorRT provider in its test

<details><summary>Claude's draft</summary>

The previous commit moved libonnxruntime_providers_tensorrt.so to
$PREFIX/lib and, along the way, dropped `import onnxruntime` from the
test. Without that import, dlopen of the TensorRT provider segfaults
(status 139 on linux-64 and linux-aarch64 CUDA 13.4). The published
1.31.0 _202 build from main crashes the same way when run without the
import, so the split is not the cause. Importing onnxruntime first loads
the CUDA/cuDNN/TensorRT dependencies the provider expects, as happens
in real use.

Resume this Claude session:
```
cd /home/mark/git/feedstock
claude --resume fdaba65c-5a43-4f26-a32f-d49b747fdbd7
```
</details>

Claude-Session: https://claude.ai/code/session_01R9WZp6UVaiXYyC2hWTNWZY

Author: hmaarrfk

recipe/recipe.yaml

@@ -391,7 +391,7 @@ outputs:
     tests:
       - script:
           # onnxruntime loads libonnxruntime_providers_shared globally before any provider.
-          - python -c "import ctypes, os, sys; [ctypes.CDLL(os.path.join(sys.prefix, 'lib', f'libonnxruntime_providers_{p}.so'), os.RTLD_GLOBAL) for p in ('shared', 'tensorrt')]"
+          - python -c "import ctypes, os, sys, onnxruntime; [ctypes.CDLL(os.path.join(sys.prefix, 'lib', f'libonnxruntime_providers_{p}.so'), os.RTLD_GLOBAL) for p in ('shared', 'tensorrt')]"
 
 about:
   homepage: https://github.com/microsoft/onnxruntime/