Also fixed a bug adding boolean when pset is already added and representation type in add_boolean
```
File "\addons\blenderbim\tool\ifc.py", line 201, in execute
IfcStore.execute_ifc_operator(self, context)
File "\addons\blenderbim\bim\ifc.py", line 336, in execute_ifc_operator
result = getattr(operator, "_execute")(context)
File "\addons\blenderbim\bim\module\model\opening.py", line 534, in _execute
data = json.loads(pset["Data"])
File "\addons\blenderbim\libs\site\packages\ifcopenshell\entity_instance.py", line 248, in __getitem__
if key < 0 or key >= len(self):
TypeError: '<' not supported between instances of 'str' and 'int'
```
replaced shapebuilder with general ifcopenshell code in `test_add_boolean` (the only place besides `test_shape_builder` where it was used), that way mathutils dependency in tests is localized only to `test_shape_builder.py`.
added test-safe option to makefile, so github workflow would ignore `test_shape_builder` tests for now and it wouldn't interrupt the entire workflow
added test that model.get_manual_booleans will ignore IfcBooleanClippings until we find some other solution to distinguish automatic and manual clippings. And some refactor along the way.
It wasn't passing the exclude_callback if attribute of the copied element wasn't a tuple, resulting in duplicated named profiles in case if you'd copy a IfcBooleanClippingResult #3810
For example, how copy_deep of IfcProductDefinitionShape with boolean clipping previously would work:
V IfcShapeRepresentation - copied with exlude_callback (part of .Representations[])
V IfcBooleanClippingResult - copied with exlude_callback (part of .Items[])
X IfcExtrudedAreaSolid - copied without exlude_callback becuase it's part of .FirstOperand (not an array) and we have IfcIShapeProfileDef duplicated
Case without booleanclippings:
V IfcShapeRepresentation - copied with exlude_callback (part of .Representations[])
V IfcExtrudedAreaSolid - copied with exlude_callback, since it's part of .Items[], IfcIShapeProfileDef not duplicated