What should logpdf multirow of observed row conditioned on the same observed row return?

[leocasarsa]

Related to #116, #181.

@riastradh-probcomp proposed for the single row logpdf:

On the other hand, we could consistently interpret cgpm.logpdf(A_r = 42) to mean log Pr[A_* = 42 | D, B_* = b_r, X_* = x_r] -- i.e., any time someone asks about an observed cell in an observed row, give the answer instead for the corresponding cell in a hypothetical row having all the other information from the observed row.

My current solution for computing logpdf_multirow(A_r=42, A_r = 42) (necessary for the relevance search) is:

1. unincorporate r if it is on the dataset
2. compute logpdf_multirow(A_r*=42 | A_r = 42, z_r=z_r*=k) * logpdf_multirow(A_r = 42, z_r=k) for all values of k and marginalize k out.
3. reincorporate r