Persistent Zero Probability for Documents Previously Classified as Outliers after Implementing reduce_outliers and _map_probabilities

Nobody has claimed this yet.

Following the application of `topic_model.reduce_outliers` to minimize outliers and the subsequent updating of topic allocations probabilities with `_map_probabilities`, it is observed that documents which were initially classified as outliers (denoted by a topic of -1) and had a probability of zero, retain this zero probability post updating. This raises a question regarding whether this is an intended consequence of the operations, or if there may be a misunderstanding of the process involved.

Start by reading the implementations and documentation for topic_model.reduce_outliers and _map_probabilities, then trace how documents with topic -1 are handled across both operations. Reproduce the reported zero-probability behavior with a small outlier dataset and establish whether it matches the documented contract; completion requires a clear answer and, if incorrect, a regression test for the expected probabilities.

Written by the indexing model from the issue text.