Fix synchronization issue originated from non-blocking stream (#740)

<!--

Thank you for contributing to cuOpt :)

Here are some guidelines to help the review process go smoothly.

Many thanks in advance for your cooperation!

Note: The pull request title will be included in the CHANGELOG.
-->


## Checklist

- [x] I am familiar with the [Contributing
Guidelines](https://github.com/NVIDIA/cuopt/blob/HEAD/CONTRIBUTING.md).
- Testing
   - [x] New or existing tests cover these changes
   - [ ] Added tests
   - [ ] Created an issue to follow-up
   - [ ] NA
- Documentation
   - [ ] The documentation is up to date with these changes
   - [ ] Added new documentation
   - [x] NA


<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **Refactor**
* Improved GPU synchronization behavior in the linear programming solver
for more reliable GPU execution.
* Switched routing data handling to a direct GPU-backed conversion
pathway for more consistent numeric types.

* **Tests**
* Updated routing tests to align with the new GPU data conversion
approach.

<sub>✏️ Tip: You can customize this high-level summary in your review
settings.</sub>
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

Author: rgsl888prabhu

cpp/src/linear_programming/utilities/cython_solve.cu

@@ -230,9 +230,12 @@ std::unique_ptr<solver_ret_t> call_solve(
 {
   raft::common::nvtx::range fun_scope("Call Solve");
 
-  cudaStream_t stream;
-  RAFT_CUDA_TRY(cudaStreamCreateWithFlags(&stream, flags));
-  const raft::handle_t handle_{stream};
+  // FIX: Use default handle constructor like CLI does, instead of explicit stream creation
+  // Original code created a non-blocking stream which causes synchronization issues with PDLP
+  // This is a workaround to fix the synchronization issues, please fix this in the future and
+  // remove this workaround. cudaStream_t stream; RAFT_CUDA_TRY(cudaStreamCreateWithFlags(&stream,
+  // flags));  // flags=cudaStreamNonBlocking const raft::handle_t handle_{stream};
+  const raft::handle_t handle_{};
 
   auto op_problem = data_model_to_optimization_problem(data_model, solver_settings, &handle_);
   solver_ret_t response;

python/cuopt/cuopt/routing/utils.py

@@ -367,7 +367,7 @@ def create_pickup_delivery_data(
             matrix[i][j] = matrix_pdf[my_ix][my_iy] / vehicle_constraints.speed
 
     pdf = pd.DataFrame(matrix)
-    matrix_df = cudf.DataFrame.from_pandas(pdf).astype("float32")
+    matrix_df = cudf.from_pandas(pdf).astype("float32")
 
     pickup_indices = cudf.Series(
         i for i in range(0, int(len(raw_order_df) / 2))

python/cuopt/cuopt/tests/routing/test_pickup_delivery.py

@@ -63,7 +63,7 @@ def data_prep(order_pdf, matrix_pdf, depot):
             matrix[i][j] = matrix_pdf[my_ix][my_iy]
 
     pdf = pd.DataFrame(matrix)
-    matrix_df = cudf.DataFrame.from_pandas(pdf).astype("float32")
+    matrix_df = cudf.from_pandas(pdf).astype("float32")
 
     # Prepare pickup and delivery indices
     delivery_indices = cudf.Series(i for i in range(0, int(len(order_df) / 2)))