Combine walk_model helper functions in the codebase?

[larryshamalama]

Description

We have two walk_models helper functions currently defined and they are very similar: one in pymc/logprob/utils.py (from AePPL's merge) and another in pymc/pytensorf.py.

pymc/pymc/logprob/utils.py

Lines 70 to 75 in a4b0581

	def walk_model(
	graphs: Iterable[TensorVariable],
	walk_past_rvs: bool = False,
	stop_at_vars: Optional[Set[TensorVariable]] = None,
	expand_fn: Callable[[TensorVariable], List[TensorVariable]] = lambda var: [],
	) -> Generator[TensorVariable, None, None]:

pymc/pymc/pytensorf.py

Lines 179 to 184 in a4b0581

	def walk_model(
	graphs: Iterable[TensorVariable],
	stop_at_vars: Optional[Set[TensorVariable]] = None,
	expand_fn: Callable[[TensorVariable], Iterable[TensorVariable]] = lambda var: [],
	) -> Generator[TensorVariable, None, None]:
	"""Walk model graphs and yield their nodes.

We can combine the two and remove one due to redundancy. Later on, we could even make it more general, depending on the use case (e.g. in #6834).

CC @ricardoV94 @shreyas3156

[ricardoV94]

They are subtly different. The walk model in logprob.utils is basically only used for rvs_to_value_vars, which is itself only used in conditional_logp. They are both less flexible/have a specific use that assumes among other things, that we are replacing things in topological order iteratively. It may make more sense to move them to logprob/basic.py and make them internal (leading underscore), to emphasize they are not general