devel/llvm13: Disable OpenMP amdgcn bit code

Due to a bug in the CMakefiles these build for aarch64 and maybe
ppc64le.  This resulted in plist breakage for aarch64.

It would be straightforward to enable them for amd64, but I have no way
to test them and the current CMake bits build them for linux targets so
I suspect they won't work.

PR:		258706
Reviewed by:	jbeich
Sponsored by:	DARPA

devel/llvm13/Makefile

@@ -174,6 +174,7 @@ LLDB_USES=		lua:53 libedit
 LLDB_VARS=		_USES_PYTHON=python:3.6+
 MLIR_DESC=		Multi-Level Intermediate Representation
 OPENMP_DESC=		Install libomp, the LLVM OpenMP runtime library
+OPENMP_CMAKE_IN=	-DLIBOMPTARGET_BUILD_AMDGCN_BCLIB=OFF
 OPENMP_PORTDOCS=	openmp
 PYCLANG_DESC=	Install python bindings to libclang
 PYCLANG_IMPLIES=	CLANG