[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/cmath

@@ -0,0 +1,85 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the CUDA Toolkit, 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) 2023 NVIDIA CORPORATION & AFFILIATES.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _CUDA_STD_CMATH
+#define _CUDA_STD_CMATH
+
+#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/__host_stdlib/math.h>
+
+#if _CCCL_COMPILER(NVHPC)
+#  include <cmath>
+#endif // _CCCL_COMPILER(NVHPC)
+
+#include <cuda/std/__cmath/abs.h>
+#include <cuda/std/__cmath/copysign.h>
+#include <cuda/std/__cmath/error_functions.h>
+#include <cuda/std/__cmath/exponential_functions.h>
+#include <cuda/std/__cmath/fdim.h>
+#include <cuda/std/__cmath/fma.h>
+#include <cuda/std/__cmath/fpclassify.h>
+#include <cuda/std/__cmath/gamma.h>
+#include <cuda/std/__cmath/hyperbolic_functions.h>
+#include <cuda/std/__cmath/hypot.h>
+#include <cuda/std/__cmath/inverse_hyperbolic_functions.h>
+#include <cuda/std/__cmath/inverse_trigonometric_functions.h>
+#include <cuda/std/__cmath/isfinite.h>
+#include <cuda/std/__cmath/isinf.h>
+#include <cuda/std/__cmath/isnan.h>
+#include <cuda/std/__cmath/isnormal.h>
+#include <cuda/std/__cmath/lerp.h>
+#include <cuda/std/__cmath/logarithms.h>
+#include <cuda/std/__cmath/min_max.h>
+#include <cuda/std/__cmath/modulo.h>
+#include <cuda/std/__cmath/nan.h>
+#include <cuda/std/__cmath/remainder.h>
+#include <cuda/std/__cmath/roots.h>
+#include <cuda/std/__cmath/rounding_functions.h>
+#include <cuda/std/__cmath/signbit.h>
+#include <cuda/std/__cmath/traits.h>
+#include <cuda/std/__cmath/trigonometric_functions.h>
+#include <cuda/std/__cstdlib/abs.h>
+#include <cuda/std/limits>
+#include <cuda/std/version>
+
+#if _CCCL_FREESTANDING()
+#  define INFINITY ::cuda::std::numeric_limits<float>::infinity()
+#  define NAN      ::cuda::std::numeric_limits<float>::quiet_NaN()
+#endif // _CCCL_FREESTANDING()
+
+#include <cuda/std/__cccl/prologue.h>
+
+_CCCL_BEGIN_NAMESPACE_CUDA_STD
+
+#if _CCCL_FREESTANDING()
+
+using double_t = double;
+using float_t  = float;
+
+#else // ^^^ _CCCL_FREESTANDING() ^^^ / vvv _CCCL_HOSTED() vvv
+
+using ::double_t;
+using ::float_t;
+
+#endif // _CCCL_HOSTED()
+
+_CCCL_END_NAMESPACE_CUDA_STD
+
+#include <cuda/std/__cccl/epilogue.h>
+
+#endif // _CUDA_STD_CMATH