Scan logprob fails when not all RVs are valued

[ricardoV94]

import aesara
import aesara.tensor as at
import aeppl

# init = at.random.beta(1, 1)
(x, y), _ = aesara.scan(
    fn=(lambda x, y: (at.random.normal(), at.random.normal())),
    outputs_info=[at.zeros(()), at.zeros(())],
    n_steps=10,
)
x.name = "x"
y.name = "y"

x_vv = x.clone()
y_vv = y.clone()

logp_dict = aeppl.factorized_joint_logprob({x: x_vv, y: y_vv})
print(len(logp_dict))  # 2

logp_dict = aeppl.factorized_joint_logprob({x: x_vv})
print(len(logp_dict))  # 0

logp_dict = aeppl.factorized_joint_logprob({y: y_vv})
print(len(logp_dict))  # 0