Fixed Bugs with Splitter in Conversation

The split_by_topic was adding the last message from the previous splitting to the current splitting output, resulting in duplicates.

I also made the code more readable.

Also moved Conversation from schema.py to text.py to avoid cyclic imports.