Fail tests gracefully for the minimal cuda build (#21391)

### Description
Several tests result in segfaults during the minimal cuda build.
Although test failures are expected due to the limitation of the minimal
cuda EP, failing gracefully would be much preferred.



### Motivation and Context
To reproduce:
1. Build ORT with:
```bash
./build.sh --build_shared_lib --use_full_protobuf --cuda_home /usr/local/cuda --cudnn_home /usr/lib/x86_64-linux-gnu/ --tensorrt_home /TensorRT-10.0.1.6 --parallel --skip_tests --skip_submodule_sync --allow_running_as_root --use_tensorrt --cmake_extra_defines onnxruntime_CUDA_MINIMAL=1
```
2. Run `onnxruntime_test_all`
```bash
...
[----------] 1 test from AllocationPlannerTest
[ RUN      ] AllocationPlannerTest.ReusedInputCrossDifferentStreams
Segmentation fault (core dumped)
```

cmake/onnxruntime_unittests.cmake

@@ -79,6 +79,9 @@ function(AddTest)
     if (onnxruntime_USE_NCCL)
       target_include_directories(${_UT_TARGET} PRIVATE ${NCCL_INCLUDE_DIRS})
     endif()
+    if(onnxruntime_CUDA_MINIMAL)
+      target_compile_definitions(${_UT_TARGET} PRIVATE -DUSE_CUDA_MINIMAL)
+    endif()
   endif()
   if (onnxruntime_USE_TENSORRT)
     # used for instantiating placeholder TRT builder to mitigate TRT library load/unload overhead

onnxruntime/test/contrib_ops/bias_dropout_op_test.cc

@@ -2,7 +2,7 @@
 // Licensed under the MIT License.
 
 // BiasDropout kernel is only implemented for CUDA/ROCM
-#if defined(USE_CUDA) || defined(USE_ROCM)
+#if (defined(USE_CUDA) && !defined(USE_CUDA_MINIMAL)) || defined(USE_ROCM)
 
 #ifdef _MSC_VER
 #pragma warning(disable : 4389)

onnxruntime/test/framework/allocation_planner_test.cc

@@ -2078,6 +2078,7 @@ TEST(AllocationPlannerTest, ReusedInputCrossDifferentStreams) {
   ASSERT_EQ(plan->allocation_plan[14].alloc_kind, AllocKind::kReuse) << "The input of reshape and gather will reuse the output of shape";
 
   int gather_count = 0;
+  ASSERT_GT(plan->execution_plan.size(), 1) << "Number of execution plans should be greater than 1";
   for (size_t i = 0; i < plan->execution_plan[1]->steps_.size(); i++) {
     if (strstr(typeid(*(plan->execution_plan[1]->steps_[i])).name(), "LaunchKernelStep")) {
       const Node* node = sess.GetSessionState().GetGraphViewer().GetNode(plan->execution_plan[1]->steps_[i]->GetNodeIndex());