Improve performance of Mesh.add method for lists

[isteinbrecher]

The common way to add objects to a Mesh is to use the add method. This allows to check that no duplicate entries exist in the mesh (if this check is really necessary can be debated). Up until now when we added a list of items, we added each element on its own via the add method in a recursive manner. This works fine as long as there are not too many entries in the list.

However, it a list with a large number of entries is added, we run into the problem, that for each element in the list, we check if this element is already in the mesh, then extend the corresponding data structure in the mesh and then do the same thing for the next item. This scales very badly due to the repeated checking for duplicate entries. A better approach is to add the whole list to the mesh and then check in the end, if we have double elements. By doing so we only have to check for duplicate entries once and the performance is much better.

This allows do directly use the add method in the creation of the space time meshes, instead of the previous hack. There we can also nicely see the performance impact, as the space time performance test with using the old add method did not finish after 5 minutes on my machine, but with the new method, we achieve basically the same performance as with the previous hack.

[davidrudlstorfer]

Thanks for the nice changes!

I have a few small suggestions

[davidrudlstorfer]

Could we simplify this construct?

For example

Suggested change

	if isinstance(item, Mesh):
	return Mesh
	elif isinstance(item, Function):
	return Function
	elif isinstance(item, BoundaryConditionBase):
	return BoundaryConditionBase
	elif isinstance(item, Material):
	return Material
	elif isinstance(item, Node):
	return Node
	elif isinstance(item, Element):
	return Element
	elif isinstance(item, GeometrySetBase):
	return GeometrySetBase
	elif isinstance(item, GeometryName):
	return GeometryName
	else:
	return type(item)
	for cls in (Mesh, Function, BoundaryConditionBase, Material, Node, Element, GeometrySetBase, GeometryName):
	if isinstance(item, cls):
	return cls
	return type(item)

[davidrudlstorfer]

Again maybe directly simplify this, for example

method_map = {
    Mesh: self.add_mesh,
    Function: self.add_function,
    BoundaryConditionBase: self.add_bc,
    Material: self.add_material,
    Node: self.add_node,
    Element: self.add_element,
    GeometrySetBase: self.add_geometry_set,
    GeometryName: self.add_geometry_name,
    list: self.add_list,
}

if base_type in method_map:
    method_map[base_type](add_item, **kwargs)
else:
    raise TypeError(f'No Mesh.add case implemented for type: "{type(add_item)}"!')

[davidrudlstorfer]

Optionally you could directly add the type hints if we add/touch functions throughout the code base

[davidrudlstorfer]

And optionally we could also directly add the args here in the docstring

[davidrudlstorfer]

Why do we create a dictionary here? Would a simple list also suffice?