Option to retrieve text instead of document ids in the topic dataset

[Pclanglais]

For some corpora it's more practical to get the actual text (especially since BERTopic works way better on shorter text, length is not really an issue).

(mostly a reminder for myself)