Advanced tutorial: embedding, alignment, and latent distribution test

[bdpedigo]

Tutorial should describe:

• spectral embeddings have a non-identifiability (orthogonal transformations in the embedded space)
• we can use the tools in the align module to circumvent this (both normal procrustes if a vertex alignment is known, or a seedless procrustes if not)
• with these tools, latent distribution test for finite samples becomes tractable

[alyakin314]

related to #177 and #414.

upd: but not the same thing, as those refers to the basic tutorial.

[alyakin314]

of some relevance: https://nbviewer.jupyter.org/url/alyakin314.github.io/assets/seedless_2.ipynb also, Ali, cep and YP have another simulated example in which it is required to use seedless to do sound inference.