Review _should_include_node implementation, that changes test nodes tags

[tatiana]

For several releases, including 1.8.x, the method _should_include_node changes the nodes of type DbtResourceType.TEST to contain the parent node labels:

astronomer-cosmos/cosmos/dbt/selector.py

Lines 387 to 407 in 74c7c7d

	def _should_include_node(self, node_id: str, node: DbtNode) -> bool:
	"""
	Checks if a single node should be included. Only runs once per node with caching."""
	logger.debug("Inspecting if the node <%s> should be included.", node_id)
	if node_id in self.visited_nodes:
	return node_id in self.selected_nodes
	self.visited_nodes.add(node_id)
	# Disclaimer: this method currently copies the tags from parent nodes to children nodes
	# that are tests. This can lead to incorrect results in graph node selectors such as reported in
	# https://github.com/astronomer/astronomer-cosmos/pull/1466
	if node.resource_type == DbtResourceType.TEST and node.depends_on and len(node.depends_on) > 0:
	node.tags = getattr(self.nodes.get(node.depends_on[0]), "tags", [])
	logger.debug(
	"The test node <%s> inherited these tags from the parent node <%s>: %s",
	node_id,
	node.depends_on[0],
	node.tags,
	)

While this was introduced in the past for some good reason - that we no longer remember - this can be very error-prone, as observed in #1466. The problem was very hard to troubleshoot, and - as the code is - we can still have other unexpected behaviours in Cosmos due to us injecting tags to nodes that didn't initially have them.

The goal of this ticket is:

1. To understand why we are changing the test nodes
2. If possible, modify this part of the code so it becomes less error-prone in the future