How `update_topics` re-computes or re-assigns probabilities?

Nobody has claimed this yet.

After getting topics, I need to split them based on internal labels. Doing so, I'll need to use `update_topics` with a custom list of topics. But `update_topics` function does not update `probabilities_` anymore. Later I'll need a probability assigned to each document in my pipeline.

While documentations [point out](https://github.com/MaartenGr/BERTopic/blob/94064441233564f29b5ac03ff5c38669067ba4e1/bertopic/_bertopic.py#L1356C6-L1356C6) that `update_topics` can get custom topics but it doesn't seem to be coherently changing all the states. Am I missing something here?

Start with the documented update_topics entry point in bertopic/_bertopic.py around line 1356, then trace how custom topics and probabilities_ are handled. Reproduce the reported sequence of getting topics, assigning a custom topic list, and requesting document probabilities. Done means the relationship between update_topics and probabilities_ is consistent or the documentation clearly states the limitation.

Written by the indexing model from the issue text.