ifcopenshell: fix geom.iterator memory growth on the OCC path #6904

On the pythonOCC path (use_python_opencascade=True) the iterator emits
SERIALIZED elements, deserialized per element by create_shape_from_serialization.
For OCC >= 7.8 this used the single-argument breptools.ReadFromString(brep_data)
overload, which pythonOCC documents as allocating and returning a new
TopoDS_Shape on every call ("increases memory"); the by-reference overload
ReadFromString(brep_data, shape) exists specifically to prevent that growth.
Use the by-reference overload, with a TypeError fallback for older bindings
that only expose the single-argument form.

Generated with the assistance of an AI coding tool.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

src/ifcopenshell-python/ifcopenshell/geom/occ_utils.py

@@ -295,8 +295,16 @@ def create_shape_from_serialization(
             ss.ReadFromString(brep_data)
             occ_shape = ss.Shape(ss.NbShapes())
         else:
+            # The single-argument breptools.ReadFromString allocates and returns
+            # a new TopoDS_Shape on every call, which pythonOCC documents as
+            # increasing memory; pass the shape by reference to reuse it instead
+            # and fall back for older bindings lacking that overload. See #6904.
             ss = BRepTools.breptools()
-            occ_shape = ss.ReadFromString(brep_data)
+            occ_shape = TopoDS.TopoDS_Shape()
+            try:
+                ss.ReadFromString(brep_data, occ_shape)
+            except TypeError:
+                occ_shape = ss.ReadFromString(brep_data)
     except BaseException as e:
         print("Error occurred parsing a shape from a string:", e)