Improve the handling of CMake deduplication (#1450)

Certain CMake functions deduplicates arguments by default. For example, if we
have two `target_link_options` with both `-Xoffload-linker -opt-A` and then
`-Xoffload-linker -opt-B`, the final link command would be `-Xoffload-linker
-opt-A -opt-B`, which is not what we want.

CMakeLists.txt

@@ -833,7 +833,7 @@ else()
   message(STATUS "Building shared RCCL library")
 endif()
 if (HAVE_KERNARG_PRELOAD)
-  target_link_options(rccl PRIVATE -Xoffload-linker -mllvm=-amdgpu-kernarg-preload-count=16)
+  target_link_options(rccl PRIVATE "SHELL:-Xoffload-linker -mllvm=-amdgpu-kernarg-preload-count=16")
 endif()
 
 if(ENABLE_MSCCLPP)