[INFRA] Import NVIDIA/CCCL upstream as optimization reference library

CCCL (CUDA C++ Core Libraries) provides:
- CUB: device/block/warp-level GPU primitives (reduce, scan, sort, topk)
- Thrust: high-level parallel algorithms (transform_reduce, sort, scan)
- libcudacxx: CUDA C++ standard library (atomics, barriers, memory)
- cudax: experimental features (memory resources, allocators)
- Tuning policies: per-SM hardware-specific algorithm parameters

Competition optimization vectors mapped to CCCL:
- Output TPS (83% weight): warp_reduce, block_reduce, device_topk
- Input TPS (14% weight): device_scan, block_load, prefetch
- Cache TPS (3% weight): prefix caching strategy patterns
- Memory (0.9 util): pooled/cached/buddy allocators

Source: https://github.com/NVIDIA/cccl (shallow clone, HEAD only)
License: Apache-2.0

cccl_upstream/libcudacxx/include/cuda/std/__latch/latch.h

@@ -0,0 +1,88 @@
+// -*- C++ -*-
+//===--------------------------- latch -----------------------------------===//
+//
+// Part of the LLVM Project, 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
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_STD___LATCH_LATCH_H
+#define _CUDA_STD___LATCH_LATCH_H
+
+#include <cuda/std/detail/__config>
+
+#if defined(_CCCL_IMPLICIT_SYSTEM_HEADER_GCC)
+#  pragma GCC system_header
+#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_CLANG)
+#  pragma clang system_header
+#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_MSVC)
+#  pragma system_header
+#endif // no system header
+
+#include <cuda/std/atomic>
+#include <cuda/std/limits>
+
+#include <cuda/std/__cccl/prologue.h>
+
+_CCCL_BEGIN_NAMESPACE_CUDA_STD
+
+template <thread_scope _Sco = thread_scope_system>
+class __latch_base
+{
+  __atomic_impl<ptrdiff_t, _Sco> __counter;
+
+public:
+  _CCCL_HOST_DEVICE_API constexpr explicit __latch_base(ptrdiff_t __expected)
+      : __counter(__expected)
+  {}
+
+  _CCCL_HIDE_FROM_ABI ~__latch_base()          = default;
+  __latch_base(const __latch_base&)            = delete;
+  __latch_base& operator=(const __latch_base&) = delete;
+
+  _CCCL_HOST_DEVICE_API inline void count_down(ptrdiff_t __update = 1)
+  {
+    _CCCL_ASSERT(__update > 0, "");
+    auto const __old = __counter.fetch_sub(__update, memory_order_release);
+    _CCCL_ASSERT(__old >= __update, "");
+    if (__old == __update)
+    {
+      __counter.notify_all();
+    }
+  }
+  _CCCL_HOST_DEVICE_API inline bool try_wait() const noexcept
+  {
+    return __counter.load(memory_order_acquire) == 0;
+  }
+  _CCCL_HOST_DEVICE_API inline void wait() const
+  {
+    while (1)
+    {
+      auto const __current = __counter.load(memory_order_acquire);
+      if (__current == 0)
+      {
+        return;
+      }
+      __counter.wait(__current, memory_order_relaxed);
+    }
+  }
+  _CCCL_HOST_DEVICE_API inline void arrive_and_wait(ptrdiff_t __update = 1)
+  {
+    count_down(__update);
+    wait();
+  }
+
+  _CCCL_HOST_DEVICE_API static constexpr ptrdiff_t max() noexcept
+  {
+    return numeric_limits<ptrdiff_t>::max();
+  }
+};
+
+using latch = __latch_base<>;
+
+_CCCL_END_NAMESPACE_CUDA_STD
+
+#include <cuda/std/__cccl/epilogue.h>
+
+#endif //_CUDA_STD___LATCH_LATCH_H