Fix build gating on optimized_portable_kernels

If we don't need optimized kernels, we don't need optimized portable kernels.
ghstack-source-id: fe2103a
ghstack-comment-id: 2755985472
Pull-Request-resolved: pytorch/executorch#9665

Author: swolchok

kernels/portable/CMakeLists.txt

@@ -66,9 +66,7 @@ gen_operators_lib(
 # Portable kernels support optional parallelization (and, in the
 # future, perhaps other performance features). If support is present,
 # produce an optimized version.
-set(BUILD_OPTIMIZED_PORTABLE_KERNELS EXECUTORCH_BUILD_PTHREADPOOL)
-
-if(BUILD_OPTIMIZED_PORTABLE_KERNELS)
+if(EXECUTORCH_BUILD_PTHREADPOOL AND EXECUTORCH_BUILD_KERNELS_OPTIMIZED)
   add_library(optimized_portable_kernels ${_portable_kernels__srcs})
   target_link_libraries(optimized_portable_kernels PRIVATE executorch)
   target_link_libraries(optimized_portable_kernels PUBLIC extension_threadpool)