Add discussion of BERT

[vanatteveldt]

We have a section on embeddings, but nothing on transformers / BERT models. This obviously needs to change

[vanatteveldt]

Concretely, I think we should add a general section in chapter 8.4 on encoder / transformer models, and add a specific section in 11.4 on BERT models for text analysis.

We might also have to review the general text about bag-of-word models to make sure it reflects that for current language models these are generally not used.