[FEA]: Implement cccl.c.parallel version of merge sort

[gevtushenko]

Is this a duplicate?

• I confirmed there appear to be no duplicate issues for this request and that I agree to the Code of Conduct

Area

Not sure

Is your feature request related to a problem? Please describe.

To implement cuda.parallel.merge_sort algorithms, we need a C version of cub::DeviceMergeSort first.

Describe the solution you'd like

We need a version of C API for merge sort, similar to what we have for reduction:

cccl/c/parallel/include/cccl/c/reduce.h

Lines 21 to 57 in 87ef1d5

	struct cccl_device_reduce_build_result_t
	{
	int cc;
	void* cubin;
	size_t cubin_size;
	CUlibrary library;
	CUkernel single_tile_kernel;
	CUkernel single_tile_second_kernel;
	CUkernel reduction_kernel;
	};
	// TODO return a union of nvtx/cuda/nvrtc errors or a string?
	extern "C" CCCL_C_API CUresult cccl_device_reduce_build(
	cccl_device_reduce_build_result_t* build,
	cccl_iterator_t d_in,
	cccl_iterator_t d_out,
	cccl_op_t op,
	cccl_value_t init,
	int cc_major,
	int cc_minor,
	const char* cub_path,
	const char* thrust_path,
	const char* libcudacxx_path,
	const char* ctk_path) noexcept;
	extern "C" CCCL_C_API CUresult cccl_device_reduce(
	cccl_device_reduce_build_result_t build,
	void* d_temp_storage,
	size_t* temp_storage_bytes,
	cccl_iterator_t d_in,
	cccl_iterator_t d_out,
	unsigned long long num_items,
	cccl_op_t op,
	cccl_value_t init,
	CUstream stream) noexcept;
	extern "C" CCCL_C_API CUresult cccl_device_reduce_cleanup(cccl_device_reduce_build_result_t* bld_ptr);

Describe alternatives you've considered

No response

Additional context

No response