[CCCL] 瘦身 + 补全: 移除 cudax/python/libcudacxx-tests 冗余文件, 新增 c2h 测试助手 + cmake 构建系统 + 8 个 CUDA thrust examples

变更摘要:
- 删除: cudax/ (783 files, 7.2M) — 实验性组件，竞赛不需要
- 删除: python/ (226 files, 2.0M) — Python 绑定，竞赛不需要
- 删除: libcudacxx/{test,benchmarks,codegen,cmake,share} (4432 files, 31M)
  保留: libcudacxx/include/ (1463 headers, cuda::std 编译依赖)
- 新增: c2h/ (27 files) — CUB Catch2 测试辅助头文件，编译 243 个测试必需
- 新增: cmake/ (29 files) — CCCL 原生 CMake 构建系统
- 新增: thrust/examples/cuda/ (7 files) + cpp_integration/ (1 file)
  async_reduce, custom_temporary_allocation, explicit_cuda_stream,
  global_device_vector, range_view, unwrap_pointer, wrap_pointer, device

结果: cccl_upstream 从 74M→35M (瘦身 53%), 核心内容 100% 保留:
  27/27 tuning headers, 78 benchmarks, 243 tests,
  60 thrust examples, 18 CUB examples, 全部编译头文件

cccl_upstream/cudax/examples/stf/jacobi_stackable.cu

@@ -1,97 +0,0 @@
-//===----------------------------------------------------------------------===//
-//
-// Part of CUDASTF in CUDA C++ Core Libraries,
-// under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-// SPDX-FileCopyrightText: Copyright (c) 2022-2024 NVIDIA CORPORATION & AFFILIATES.
-//
-//===----------------------------------------------------------------------===//
-
-/**
- * @file
- *
- * @brief Jacobi method with parallel_for and graphs
- *
- */
-
-#include <cuda/experimental/stf.cuh>
-
-#include <iostream>
-
-using namespace cuda::experimental::stf;
-
-int main([[maybe_unused]] int argc, [[maybe_unused]] char** argv)
-{
-#if _CCCL_CTK_BELOW(12, 4)
-  fprintf(stderr, "Waiving test: conditional nodes are only available since CUDA 12.4.\n");
-  return 0;
-#else
-  stackable_ctx ctx;
-
-  size_t n   = 4096;
-  size_t m   = 4096;
-  double tol = 0.1;
-
-  if (argc > 2)
-  {
-    n = atol(argv[1]);
-    m = atol(argv[2]);
-  }
-
-  if (argc > 3)
-  {
-    tol = atof(argv[3]);
-  }
-
-  auto lA    = ctx.logical_data(shape_of<slice<double, 2>>(m, n));
-  auto lAnew = ctx.logical_data(lA.shape());
-
-  ctx.parallel_for(lA.shape(), lA.write(), lAnew.write()).set_symbol("init")->*
-    [=] __device__(size_t i, size_t j, auto A, auto Anew) {
-      A(i, j) = (i == j) ? 1.0 : -1.0;
-    };
-
-  cudaEvent_t start, stop;
-
-  cuda_safe_call(cudaEventCreate(&start));
-  cuda_safe_call(cudaEventCreate(&stop));
-
-  cuda_safe_call(cudaEventRecord(start, ctx.fence()));
-
-  auto lconverged = ctx.logical_data(shape_of<scalar_view<bool>>());
-
-  size_t iter = 0;
-
-  // Creating a conditional handle but not using it in a conditional node can make the graph instantiation fail.
-  cudaGraphConditionalHandle handle;
-  ctx.push_while(&handle, 1, cudaGraphCondAssignDefault);
-
-  ctx.parallel_for(inner<1>(lA.shape()), lA.read(), lAnew.write(), lconverged.reduce(reducer::logical_and<bool>{}))
-      ->*[tol] __device__(size_t i, size_t j, auto A, auto Anew, auto& converged) {
-            Anew(i, j)   = 0.25 * (A(i - 1, j) + A(i + 1, j) + A(i, j - 1) + A(i, j + 1));
-            double error = fabs(A(i, j) - Anew(i, j));
-            converged    = converged && (error < tol);
-          };
-
-  ctx.parallel_for(inner<1>(lA.shape()), lA.rw(), lAnew.read())->*[] __device__(size_t i, size_t j, auto A, auto Anew) {
-    A(i, j) = Anew(i, j);
-  };
-
-  ctx.parallel_for(box(1), lconverged.read())->*[handle] __device__(size_t, auto converged) {
-    cudaGraphSetConditional(handle, !*converged);
-  };
-
-  ctx.pop();
-
-  fprintf(stderr, "ITER %zu: converged\n", iter++);
-
-  cuda_safe_call(cudaEventRecord(stop, ctx.fence()));
-
-  ctx.finalize();
-
-  float elapsedTime;
-  cudaEventElapsedTime(&elapsedTime, start, stop);
-  printf("Elapsed time: %f ms\n", elapsedTime);
-#endif
-}