Add docs on RAG citation modes

[mrmer1]

This pull request introduces a new section titled "Citation modes" to the documentation for Retrieval Augmented Generation (RAG) in streaming mode. It explains the two citation modes available: fast citations and accurate citations.

• Accurate citations: The model generates its answer first, then provides citations after the entire response is generated. This ensures precise alignment of citation indices with the final text segments but may increase latency. The example code demonstrates how to use accurate citations with the citation_quality="accurate" argument in the API call.
• Fast citations: The model generates citations inline as the response is produced, providing immediate traceability but with slightly less precision. The example code shows how to use fast citations with the citation_quality="fast" argument in the API call.

The section also emphasizes the importance of inline citations for user verification, as RAG does not guarantee accuracy or eliminate the risk of hallucination.

[github-actions]

🌿 Preview your docs: https://cohere-preview-65ee0aef-84e4-4858-a693-c1be5f806a51.docs.buildwithfern.com

[github-actions]

🌿 Preview your docs: https://cohere-preview-eaf17d66-84a2-4bb4-9d52-11acacf47b14.docs.buildwithfern.com