emitMappedTensorAccess: use isl::multi_aff::set_range_tuple_id

This removes a use of isl::dim_type, which will not be exported
in mainline isl.

Author: Sven Verdoolaege

tc/core/polyhedral/cuda/codegen.cc

@@ -674,7 +674,7 @@ void emitMappedTensorAccess(
   auto access =
       makeMultiAffAccess(tensorId, subscripts, context); // MA :: D -> O
   auto promotion = promotionInfo.group->promotion(); // MA :: [S -> O] -> P
-  promotion = promotion.set_tuple_id(isl::dim_type::out, promotionInfo.groupId);
+  promotion = promotion.set_range_tuple_id(promotionInfo.groupId);
   auto iteratorMap = context.iteratorMap(); // PMA :: A -> D
   auto schedule =
       isl::map::from_union_map(promotionInfo.outerSchedule.intersect_domain(