Replace deprecated cuda.cudart with cuda.bindings.runtime. (#8790)

Fixes #8789 

### Description

Replaces the deprecated `cuda.cudart` import in `trt_compiler.py` with
`cuda.bindings.runtime`, which is the current API introduced in
`cuda-bindings` >= 12.6 (pulled in by PyTorch 2.10+).

### Types of changes
<!--- Put an `x` in all the boxes that apply, and remove the not
applicable items -->
- [x] Non-breaking change (fix or new feature that would not break
existing functionality).
- [ ] Breaking change (fix or new feature that would cause existing
functionality to change).
- [ ] New tests added to cover the changes.
- [ ] Integration tests passed locally by running `./runtests.sh -f -u
--net --coverage`.
- [ ] Quick tests passed locally by running `./runtests.sh --quick
--unittests --disttests`.
- [ ] In-line docstrings updated.
- [ ] Documentation updated, tested `make html` command in the `docs/`
folder.

---------

Signed-off-by: ytl0623 <david89062388@gmail.com>
Co-authored-by: Eric Kerfoot <17726042+ericspod@users.noreply.github.com>

Author: ytl0623

monai/networks/trt_compiler.py

@@ -85,11 +85,22 @@ def cuassert(cuda_ret):
     """
     Error reporting method for CUDA calls.
     Args:
-     cuda_ret: CUDA return code.
+        cuda_ret: Tuple returned by CUDA runtime calls, where the first element
+            is a cudaError_t enum and subsequent elements are return values.
+
+    Raises:
+        RuntimeError: If the CUDA call returned an error.
     """
     err = cuda_ret[0]
-    if err != 0:
-        raise RuntimeError(f"CUDA ERROR: {err}")
+    if err.value != 0:
+        err_msg = f"CUDA ERROR: {err.value}"
+        try:
+            _, err_name = cudart.cudaGetErrorName(err)
+            _, err_str = cudart.cudaGetErrorString(err)
+            err_msg = f"CUDA ERROR {err.value}: {err_name} — {err_str}"
+        except Exception as e:
+            get_logger("monai.networks.trt_compiler").debug(f"Failed to retrieve CUDA error details: {e}")
+        raise RuntimeError(err_msg)
     if len(cuda_ret) > 1:
         return cuda_ret[1]
     return None