The main goal of this PR is to remove two lines:

self.__class__ = type(f"{schema.kind}InfrahubNode", (self.__class__,), {})

and:

self.__class__ = type(f"{schema.kind}InfrahubNodeSync", (self.__class__,), {})

What these lines do are to create a dynamic Python class for each object we define with the SDK when we create an InfrahubNode or InfrahubNodeSync object.

The main need for the creation of a new object came from the generate_relationship_property() function that created a setter/getter for cardinality=one relationships on the new dynamically created class.

This allowed you to override the data directly on a relationship of cardinality one at the attribute level. In pseudo code if we had a node that looked like this:

class InfraInterface:
    name: Attribute
    device: Relationship

Then if we had an instance of this as an InfrahubNode object:

interface.name.value = "Ethernet1" # Correctly sets the name attribute
interface.name = "Ethernet1" # Previously this would overwrite the name attribute and instead set "name" as a simple `str`, which wouldn't work
interface.device = "453a7da5-2c71-4cac-b96f-e00b787e0436" # this uses the attribute setter and assigns a value under the relationship

With the new implementation both attributes and cardinality=one relationships have both a setter and a getter, so the example above with interface.name = "Ethernet1". Using the same type of logic we'd be able to add a setter to cardinality=many relationships as well, we'd only have to decide how it would work if it should add a relationship or replace all previous ones.

This PR changes the way we would have to manage typing, there are a few variations we could do for this an example can be seen in the two current commits of this PR within the file query_groups.py on the line self.previous_members = group.members.peers the first commit only updates the rules, the second commit removes the ignore rules but could make the code a bit harder to read.

Test pipeline within the Infrahub project: opsmill/infrahub#6492

One change with this is that when using a Python REPL and using dir(my_infrahub_node_object) the name of the attributes and relationships won't be seen as attribute objects.

I'm also thinking if we should have infrahubnode_ as a reserved prefix so that we can add public functions using that name without having to fear future conflicts.