fix linker errors about multiply defined symbols in STF

[ericniebler]

Description

my pr's are failing CI with errors like this:

/usr/bin/ld: cudax/CMakeFiles/cudax.cpp17.headers.basic.stf.dir/./headers/cudax.cpp17.headers.basic.stf/cuda/experimental/__stf/places/exec/host/callback_queues.cuh.cu.o: in function `cuda::experimental::stf::callback_completion_kernel(int*)':
  tmpxft_00006f23_00000000-6_callback_queues.cuh.compute_80.cudafe1.cpp:(.text+0x10): multiple definition of `cuda::experimental::stf::callback_completion_kernel(int*)'; cudax/CMakeFiles/cudax.cpp17.headers.basic.stf.dir/headers/cudax.cpp17.headers.basic.stf/cuda/experimental/__stf/places/exec/host/callback_queues.cuh.cu.o:tmpxft_00006f23_00000000-6_callback_queues.cuh.compute_80.cudafe1.cpp:(.text+0x10): first defined here
  /usr/bin/ld: cudax/CMakeFiles/cudax.cpp17.headers.basic.stf.dir/./headers/cudax.cpp17.headers.basic.stf/cuda/experimental/__stf/places/exec/host/callback_queues.cuh.cu.o: in function `cuda::experimental::stf::cudaCallbackSetStatus(int, void*)':
  tmpxft_00006f23_00000000-6_callback_queues.cuh.compute_80.cudafe1.cpp:(.text+0xc0): multiple definition of `cuda::experimental::stf::cudaCallbackSetStatus(int, void*)'; cudax/CMakeFiles/cudax.cpp17.headers.basic.stf.dir/headers/cudax.cpp17.headers.basic.stf/cuda/experimental/__stf/places/exec/host/callback_queues.cuh.cu.o:tmpxft_00006f23_00000000-6_callback_queues.cuh.compute_80.cudafe1.cpp:(.text+0xc0): first defined here
  /usr/bin/ld: cudax/CMakeFiles/cudax.cpp17.headers.basic.stf.dir/./headers/cudax.cpp17.headers.basic.stf/cuda/experimental/__stf/places/exec/host/callback_queues.cuh.cu.o: in function `cuda::experimental::stf::cudaCallbackGetStatus(int*, void**)':
  tmpxft_00006f23_00000000-6_callback_queues.cuh.compute_80.cudafe1.cpp:(.text+0x150): multiple definition of `cuda::experimental::stf::cudaCallbackGetStatus(int*, void**)'; cudax/CMakeFiles/cudax.cpp17.headers.basic.stf.dir/headers/cudax.cpp17.headers.basic.stf/cuda/experimental/__stf/places/exec/host/callback_queues.cuh.cu.o:tmpxft_00006f23_00000000-6_callback_queues.cuh.compute_80.cudafe1.cpp:(.text+0x150): first defined here
  /usr/bin/ld: cudax/CMakeFiles/cudax.cpp17.headers.basic.stf.dir/./headers/cudax.cpp17.headers.basic.stf/cuda/experimental/__stf/places/exec/host/callback_queues.cuh.cu.o: in function `__device_stub__ZN4cuda12experimental3stf26callback_completion_kernelEPi(int*)':
  tmpxft_00006f23_00000000-6_callback_queues.cuh.compute_80.cudafe1.cpp:(.text+0x1f0): multiple definition of `__device_stub__ZN4cuda12experimental3stf26callback_completion_kernelEPi(int*)'; cudax/CMakeFiles/cudax.cpp17.headers.basic.stf.dir/headers/cudax.cpp17.headers.basic.stf/cuda/experimental/__stf/places/exec/host/callback_queues.cuh.cu.o:tmpxft_00006f23_00000000-6_callback_queues.cuh.compute_80.cudafe1.cpp:(.text+0x1f0): first defined here
  collect2: error: ld returned 1 exit status

this is caused by function definition in headers without inline. there's also a multiply-defined symbols error about a cuda kernel. i'm not sure what is the right way to handle kernels in headers, so i just made it a template. i would be happy to learn the right way to do this.

[github-actions]

🟩 CI finished in 36m 21s: Pass: 100%/54 | Total: 4h 02m | Avg: 4m 29s | Max: 17m 37s | Hits: 89%/224

• 🟩 cudax: Pass: 100%/54 | Total: 4h 02m | Avg: 4m 29s | Max: 17m 37s | Hits: 89%/224

  🟩 cpu
    🟩 amd64              Pass: 100%/50  | Total:  3h 51m | Avg:  4m 38s | Max: 17m 37s | Hits:  89%/224   
    🟩 arm64              Pass: 100%/4   | Total: 11m 03s | Avg:  2m 45s | Max:  3m 23s
  🟩 ctk
    🟩 12.0               Pass: 100%/19  | Total:  1h 25m | Avg:  4m 30s | Max: 17m 11s | Hits:  89%/112   
    🟩 12.5               Pass: 100%/2   | Total:  9m 30s | Avg:  4m 45s | Max:  4m 52s
    🟩 12.6               Pass: 100%/33  | Total:  2h 27m | Avg:  4m 28s | Max: 17m 37s | Hits:  89%/112   
  🟩 cudacxx
    🟩 nvcc12.0           Pass: 100%/19  | Total:  1h 25m | Avg:  4m 30s | Max: 17m 11s | Hits:  89%/112   
    🟩 nvcc12.5           Pass: 100%/2   | Total:  9m 30s | Avg:  4m 45s | Max:  4m 52s
    🟩 nvcc12.6           Pass: 100%/33  | Total:  2h 27m | Avg:  4m 28s | Max: 17m 37s | Hits:  89%/112   
  🟩 cudacxx_family
    🟩 nvcc               Pass: 100%/54  | Total:  4h 02m | Avg:  4m 29s | Max: 17m 37s | Hits:  89%/224   
  🟩 cxx
    🟩 Clang9             Pass: 100%/2   | Total:  6m 43s | Avg:  3m 21s | Max:  3m 41s
    🟩 Clang10            Pass: 100%/2   | Total:  6m 07s | Avg:  3m 03s | Max:  3m 20s
    🟩 Clang11            Pass: 100%/4   | Total: 11m 30s | Avg:  2m 52s | Max:  3m 08s
    🟩 Clang12            Pass: 100%/4   | Total: 12m 07s | Avg:  3m 01s | Max:  3m 22s
    🟩 Clang13            Pass: 100%/4   | Total: 11m 51s | Avg:  2m 57s | Max:  3m 09s
    🟩 Clang14            Pass: 100%/4   | Total: 24m 49s | Avg:  6m 12s | Max: 15m 34s
    🟩 Clang15            Pass: 100%/2   | Total:  6m 26s | Avg:  3m 13s | Max:  3m 22s
    🟩 Clang16            Pass: 100%/4   | Total: 12m 17s | Avg:  3m 04s | Max:  3m 23s
    🟩 Clang17            Pass: 100%/2   | Total:  6m 13s | Avg:  3m 06s | Max:  3m 13s
    🟩 Clang18            Pass: 100%/2   | Total: 20m 03s | Avg: 10m 01s | Max: 16m 57s
    🟩 GCC9               Pass: 100%/2   | Total:  6m 06s | Avg:  3m 03s | Max:  3m 05s
    🟩 GCC10              Pass: 100%/4   | Total: 11m 25s | Avg:  2m 51s | Max:  3m 08s
    🟩 GCC11              Pass: 100%/4   | Total: 12m 08s | Avg:  3m 02s | Max:  3m 13s
    🟩 GCC12              Pass: 100%/7   | Total:  1h 03m | Avg:  9m 02s | Max: 17m 37s
    🟩 GCC13              Pass: 100%/3   | Total:  7m 49s | Avg:  2m 36s | Max:  2m 39s
    🟩 MSVC14.36          Pass: 100%/1   | Total:  6m 46s | Avg:  6m 46s | Max:  6m 46s | Hits:  89%/112   
    🟩 MSVC14.39          Pass: 100%/1   | Total:  7m 40s | Avg:  7m 40s | Max:  7m 40s | Hits:  89%/112   
    🟩 NVHPC24.7          Pass: 100%/2   | Total:  9m 30s | Avg:  4m 45s | Max:  4m 52s
  🟩 cxx_family
    🟩 Clang              Pass: 100%/30  | Total:  1h 58m | Avg:  3m 56s | Max: 16m 57s
    🟩 GCC                Pass: 100%/20  | Total:  1h 40m | Avg:  5m 02s | Max: 17m 37s
    🟩 MSVC               Pass: 100%/2   | Total: 14m 26s | Avg:  7m 13s | Max:  7m 40s | Hits:  89%/224   
    🟩 NVHPC              Pass: 100%/2   | Total:  9m 30s | Avg:  4m 45s | Max:  4m 52s
  🟩 gpu
    🟩 v100               Pass: 100%/54  | Total:  4h 02m | Avg:  4m 29s | Max: 17m 37s | Hits:  89%/224   
  🟩 jobs
    🟩 Build              Pass: 100%/49  | Total:  2h 39m | Avg:  3m 15s | Max:  7m 40s | Hits:  89%/224   
    🟩 Test               Pass: 100%/5   | Total:  1h 23m | Avg: 16m 41s | Max: 17m 37s
  🟩 sm
    🟩 90                 Pass: 100%/1   | Total:  2m 42s | Avg:  2m 42s | Max:  2m 42s
    🟩 90a                Pass: 100%/1   | Total:  2m 39s | Avg:  2m 39s | Max:  2m 39s
  🟩 std
    🟩 17                 Pass: 100%/29  | Total:  1h 55m | Avg:  3m 59s | Max: 17m 11s
    🟩 20                 Pass: 100%/25  | Total:  2h 06m | Avg:  5m 04s | Max: 17m 37s | Hits:  89%/224   
  

👃 Inspect Changes

Modifications in project?

	Project
	CCCL Infrastructure
	libcu++
	CUB
	Thrust
+/-	CUDA Experimental
	pycuda
	CCCL C Parallel Library

Modifications in project or dependencies?

	Project
	CCCL Infrastructure
	libcu++
	CUB
	Thrust
+/-	CUDA Experimental
	pycuda
	CCCL C Parallel Library

🏃‍ Runner counts (total jobs: 54)

#	Runner
43	linux-amd64-cpu16
5	linux-amd64-gpu-v100-latest-1
4	linux-arm64-cpu16
2	windows-amd64-cpu16