Setup markdown and test PPL comparisons

[MansMeg]

We should write down how to make comparisons between PPLs so that we test that the posteriord is the same (up to a proportionality constant).

Preliminary idea is to use log densities evaluated at ref draws (or random draws) and the gradients of the log density. We will store this in posteriordb?

[MansMeg]

• See if we can get PyMC and theoretical model to agree
• Add comparison also with stan log density
• Think about potential hypothesis test to compare two empirical distributions
• Jakob write up a summary in markdown

[MansMeg]

Jakob will write up a notebook with a suggestion on test until the 16th of September.

[MansMeg]

Jakob has written a draft with a suggestion for the evaluation pipeline. Jakob has added sampling step when ref draws are missing.
Måns will read, and then we will discuss with Bob and Aki. Måns setup a decision.
See #261