Set default updates for all graph RandomVariables in compile_pymc

[ricardoV94]

This PR makes compile_pymc set default_updates for any RandomVariables found in the graph. Previously this logic was limited to NoDistribution to accommodate the special case of pm.Simulator.

With this change the following code now produces the "expected" behavior (i.e., different draws of x):

with pm.Model() as m:
  pm.Deterministic("x", pm.Normal.dist())
  prior = pm.sample_prior_predictive()

With this change we may also consider removing this responsibility from pm.Distribution which was doing this for registered (named) variables, which should facilitate #5308

pymc/pymc/distributions/distribution.py

Lines 366 to 381 in ea776d6

	rng = kwargs.pop("rng", None)
	if (
	rv_out.owner
	and isinstance(rv_out.owner.op, RandomVariable)
	and isinstance(rng, RandomStateSharedVariable)
	and not getattr(rng, "default_update", None)
	):
	# This tells `aesara.function` that the shared RNG variable
	# is mutable, which--in turn--tells the `FunctionGraph`
	# `Supervisor` feature to allow in-place updates on the variable.
	# Without it, the `RandomVariable`s could not be optimized to allow
	# in-place RNG updates, forcing all sample results from compiled
	# functions to be the same on repeated evaluations.
	new_rng = rv_out.owner.outputs[0]
	rv_out.update = (rng, new_rng)
	rng.default_update = new_rng

---

This is likely to be all reevaluated soon, as Aesara is in the process of refining the API for seeding of RandomVariables

[codecov]

Codecov Report

Merging #5442 (42229ec) into main (4fedb60) will decrease coverage by 0.02%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##             main    #5442      +/-   ##
==========================================
- Coverage   81.41%   81.39%   -0.03%     
==========================================
  Files          82       82              
  Lines       14213    14212       -1     
==========================================
- Hits        11572    11568       -4     
- Misses       2641     2644       +3     

Impacted Files	Coverage Δ
pymc/aesaraf.py	90.17% <ø> (-0.03%)	⬇️
pymc/parallel_sampling.py	86.71% <0.00%> (-1.00%)	⬇️