Build LLVMAarch64CodeGen if CMAKE_OSX_ARCHITECTURES is arm64. (#4637)

I'm trying to make arm64 Triton builds work. I've found that on my
machine, everything works if I make this change and build with
CMAKE_OSX_ARCHITECTURES=arm64.
    
I'm not totally sure why this change is needed. From talking to chatgpt,
it should be sufficient to run the build with `arch -arm64`; that should
set CMAKE_SYSTEM_PROCESSOR. But...it doesn't. And you are not allowed to
override this variable from the command-line.
    
Anyway this seems to work, and who among us can say they truly
understand cmake?

CMakeLists.txt

@@ -220,7 +220,8 @@ if(TRITON_BUILD_PYTHON_MODULE)
 
   )
   if(CMAKE_SYSTEM_PROCESSOR MATCHES "aarch64" OR # Linux arm64
-     CMAKE_SYSTEM_PROCESSOR MATCHES "arm64")     # macOS arm64
+     CMAKE_SYSTEM_PROCESSOR MATCHES "arm64" OR # macOS arm64
+     CMAKE_OSX_ARCHITECTURES MATCHES "arm64")  # also macOS arm64
       list(APPEND TRITON_LIBRARIES
           LLVMAArch64CodeGen
           LLVMAArch64AsmParser