[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/places/CMakeLists.txt

@@ -1,30 +0,0 @@
-set(places_example_sources thrust_device_data_place_allocator.cu)
-
-## cudax_add_places_example
-#
-# Add a places example executable and register it with ctest.
-#
-# target_name_var: Variable name to overwrite with the name of the example
-#   target. Useful for modifying the example/target after creation.
-# source: The source file for the example.
-#
-function(cudax_add_places_example target_name_var source)
-  get_filename_component(filename ${source} NAME_WE)
-
-  set(example_target cudax.example.places.${filename})
-
-  cccl_add_executable(${example_target} SOURCES ${source} ADD_CTEST)
-  cudax_places_configure_target(${example_target})
-  target_link_libraries(
-    ${example_target}
-    PRIVATE #
-      cudax.compiler_interface
-      cudax.examples.thrust
-  )
-
-  set(${target_name_var} ${example_target} PARENT_SCOPE)
-endfunction()
-
-foreach (source IN LISTS places_example_sources)
-  cudax_add_places_example(example_target "${source}")
-endforeach()

cccl_upstream/cudax/examples/places/thrust_device_data_place_allocator.cu

@@ -1,117 +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) 2026 NVIDIA CORPORATION & AFFILIATES.
-//
-//===----------------------------------------------------------------------===//
-
-/**
- * @file
- *
- * @brief Thrust device_vector with an allocator backed by a data_place.
- *
- * Wraps data_place::allocate/deallocate as a thrust::mr::memory_resource,
- * then uses thrust::mr::allocator to create a compatible allocator.
- * Storage is allocated via data_place (device, composite/VMM, or other
- * place types). The same Thrust code works unchanged for single-device,
- * multi-device (VMM), or green-context placement.
- */
-
-#include <thrust/copy.h>
-#include <thrust/device_ptr.h>
-#include <thrust/device_vector.h>
-#include <thrust/execution_policy.h>
-#include <thrust/host_vector.h>
-#include <thrust/iterator/counting_iterator.h>
-#include <thrust/mr/allocator.h>
-#include <thrust/mr/memory_resource.h>
-#include <thrust/transform.h>
-
-#include <cuda/experimental/__places/partitions/blocked_partition.cuh>
-
-#include <cstdio>
-
-using namespace cuda::experimental::places;
-
-// Minimal adapter: data_place is STF's abstraction; Thrust expects a
-// memory_resource. This class bridges the two. The resource must outlive
-// any vectors/allocators that use it.
-class data_place_memory_resource final : public thrust::mr::memory_resource<thrust::device_ptr<void>>
-{
-public:
-  explicit data_place_memory_resource(const data_place& place)
-      : place_(place)
-  {}
-
-  pointer do_allocate(std::size_t bytes, std::size_t /*alignment*/) override
-  {
-    // A memory resource hands out untyped bytes, so declare the geometry
-    // explicitly as a flat byte array: composite places distribute it with
-    // byte granularity (equivalent for every other place type).
-    void* raw = place_.allocate_nd(dim4(bytes), 1);
-    return thrust::device_ptr<void>(raw);
-  }
-
-  void do_deallocate(pointer p, std::size_t bytes, std::size_t /*alignment*/) override
-  {
-    place_.deallocate(p.get(), bytes);
-  }
-
-  __host__ __device__ bool do_is_equal(const memory_resource& other) const noexcept override
-  {
-#if defined(__CUDA_ARCH__)
-    (void) other;
-    return false;
-#else
-    auto* o = dynamic_cast<const data_place_memory_resource*>(&other);
-    return o && place_ == o->place_;
-#endif
-  }
-
-private:
-  data_place place_;
-};
-
-template <typename T>
-using data_place_allocator = thrust::mr::allocator<T, data_place_memory_resource>;
-
-bool run_with_place(const data_place& place, const char* label)
-{
-  const size_t n = 1024 * 1024;
-
-  data_place_memory_resource memres(place);
-  data_place_allocator<double> alloc(&memres);
-  thrust::device_vector<double, data_place_allocator<double>> d_vec(n, 0.0, alloc);
-
-  thrust::transform(
-    thrust::device,
-    thrust::counting_iterator<size_t>(0),
-    thrust::counting_iterator<size_t>(n),
-    d_vec.begin(),
-    [] __device__(size_t i) {
-      return 2.0 * static_cast<double>(i);
-    });
-
-  thrust::host_vector<double> h_sample(4);
-  thrust::copy(d_vec.begin(), d_vec.begin() + 4, h_sample.begin());
-
-  bool ok = (h_sample[0] == 0.0 && h_sample[1] == 2.0 && h_sample[2] == 4.0 && h_sample[3] == 6.0);
-  printf(
-    "thrust_device_data_place_allocator: %s (%s): %s\n", label, place.to_string().c_str(), ok ? "PASSED" : "FAILED");
-  return ok;
-}
-
-int main()
-{
-  bool all_ok = true;
-
-  all_ok &= run_with_place(data_place::device(0), "device(0)");
-
-  all_ok &= run_with_place(data_place::composite(blocked_partition(), exec_place::all_devices()),
-                           "composite(blocked_partition, all_devices)");
-
-  return all_ok ? 0 : 1;
-}