Use more readable ignore logprob

[ricardoV94]

This PR replaces uses of assign_custom_measurable_outputs by the more readable ignore_logprob.

It also refactors a utility to ignore the logprob of multiple (potentially nested) variables consistently.

This PR is a spinoff of #6529

[codecov]

Codecov Report

Merging #6596 (b67d128) into main (457421b) will decrease coverage by 0.50%.
The diff coverage is 100.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #6596      +/-   ##
==========================================
- Coverage   92.03%   91.54%   -0.50%     
==========================================
  Files          92       92              
  Lines       15539    15541       +2     
==========================================
- Hits        14302    14227      -75     
- Misses       1237     1314      +77     

Impacted Files	Coverage Δ
pymc/logprob/censoring.py	100.00% <100.00%> (ø)
pymc/logprob/cumsum.py	100.00% <100.00%> (ø)
pymc/logprob/mixture.py	98.32% <100.00%> (-0.03%)	⬇️
pymc/logprob/tensor.py	81.35% <100.00%> (-1.05%)	⬇️
pymc/logprob/transforms.py	96.35% <100.00%> (ø)
pymc/logprob/utils.py	100.00% <100.00%> (ø)

... and 4 files with indirect coverage changes

[Armavica]

Is there a reason to not inline assign_custom_measurable_outputs inside ignore_logprob since I believe it is now only used there?

[ricardoV94]

Is there a reason to not inline assign_custom_measurable_outputs inside ignore_logprob since I believe it is now only used there?

That function is a bit more powerful. It can be used to toggle some outputs on and off as measurable.

Although it's not being used anywhere else I would leave it separately until it's proven not ever neeeded

[larryshamalama]

This PR has already been merged, but I still had some questions here and there 😅

[larryshamalama]

Would it be good to add a test for ignore_logprob_multiple_vars?

[ricardoV94]

The functionality is a core requirement of stack_logprob, and it's covered in this test, when it was first introduced:

pymc/tests/logprob/test_tensor.py

Lines 134 to 171 in c709abf

	@pytest.mark.parametrize("reverse", (False, True))
	def test_measurable_make_vector_interdependent(reverse):
	"""Test that we can obtain a proper graph when stacked RVs depend on each other"""
	x = pt.random.normal(name="x")
	y_rvs = []
	prev_rv = x
	for i in range(3):
	next_rv = pt.random.normal(prev_rv + 1, name=f"y{i}")
	y_rvs.append(next_rv)
	prev_rv = next_rv
	if reverse:
	y_rvs = y_rvs[::-1]
	ys = pt.stack(y_rvs)
	ys.name = "ys"
	x_vv = x.clone()
	ys_vv = ys.clone()
	logp = joint_logprob({x: x_vv, ys: ys_vv})
	assert_no_rvs(logp)
	y0_vv = y_rvs[0].clone()
	y1_vv = y_rvs[1].clone()
	y2_vv = y_rvs[2].clone()
	ref_logp = joint_logprob({x: x_vv, y_rvs[0]: y0_vv, y_rvs[1]: y1_vv, y_rvs[2]: y2_vv})
	rng = np.random.default_rng()
	x_vv_test = rng.normal()
	ys_vv_test = rng.normal(size=3)
	np.testing.assert_allclose(
	logp.eval({x_vv: x_vv_test, ys_vv: ys_vv_test}),
	ref_logp.eval(
	{x_vv: x_vv_test, y0_vv: ys_vv_test[0], y1_vv: ys_vv_test[1], y2_vv: ys_vv_test[2]}
	),
	)

#6342

This PR just refactored it out, because it proved useful elesewhere #6529

---

Having said that, we can definitely add a targeted test.