Create example notebook for tokenizing catalog data

We may choose to just do this for images for simplicity, or we could extend to multiple modalities if we wanted to overachieve. But largely this should just provide a template for the typical machinery needed to perform tokenization, which itself is an inference style workflow rather than a pure map_partitions style operation.

Start by locating existing example notebooks and the catalog-processing entry points, then trace how tokenization fits an inference-style workflow rather than a map_partitions operation. Create a notebook that serves as a reusable tokenization template, using images as the initial scope unless the project guidance expands it to multiple modalities; done means the workflow is clear and runnable.

Written by the indexing model from the issue text.