Use-case vignettes for tidyllm

[edubruell]

Description: Develop three use-case-oriented vignettes to demonstrate how tidyllm can be used in real-world scenarios. The vignettes will cover the following:

1. Explain how to query PDF documents and return structured answers using LLMs. Include examples with JSON outputs and reasonable post-processing
2. Show an use-case of automatically labeling images with multimodal models.
3. Demonstrate synthetic data generation for questionnaires, witth large language model reponses

[edubruell]

We also need a vignette on embeddings and how to use them.

[edubruell]

Did the last of these for now. Another batch on extracting data from historical documents and working with embedding models will come sometime in November.