Fix llvm by manually setting backend in KernelFunction

pystencils/kernelcreation.py

@@ -270,7 +270,7 @@ def create_staggered_kernel(staggered_field, expressions, subexpressions=(), tar
     return ast
 
 
-def make_python_function(kernel_function_node, backend='cpu', argument_dict=None):
+def make_python_function(kernel_function_node, backend='cpu', argument_dict={}):
     """
     A generic version of the {cuda,cpu,llvm}jit.make_python_function