Investigate how LLM chat interfaces can be used for data consumers

LLMs and chat bots are fast becoming an alternative way of interfacing end-users with complex interfaces. Apache Superset has [recently](https://github.com/apache/superset/pull/41205) added an extension api for rendering chat interfaces within Superset, a preview can be found below:

https://private-user-images.githubusercontent.com/70410625/610003588-9c76b583-0407-471f-8c07-e34736dcfeb7.mov?jwt=eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3ODIzOTYxNjAsIm5iZiI6MTc4MjM5NTg2MCwicGF0aCI6Ii83MDQxMDYyNS82MTAwMDM1ODgtOWM3NmI1ODMtMDQwNy00NzFmLThjMDctZTM0NzM2ZGNmZWI3Lm1vdj9YLUFtei1BbGdvcml0aG09QVdTNC1ITUFDLVNIQTI1NiZYLUFtei1DcmVkZW50aWFsPUFLSUFWQ09EWUxTQTUzUFFLNFpBJTJGMjAyNjA2MjUlMkZ1cy1lYXN0LTElMkZzMyUyRmF3czRfcmVxdWVzdCZYLUFtei1EYXRlPTIwMjYwNjI1VDEzNTc0MFomWC1BbXotRXhwaXJlcz0zMDAmWC1BbXotU2lnbmF0dXJlPTcwYTcxMzMxNmVmMTEyZDQ3NjY3ODAxZDkxYzRmZjZhYmY3NWUwYzc5YjAzOGIxNjEyZWYyZGRhNmExZTNhNzMmWC1BbXotU2lnbmVkSGVhZGVycz1ob3N0JnJlc3BvbnNlLWNvbnRlbnQtdHlwZT12aWRlbyUyRnF1aWNrdGltZSJ9.d-YEOXQ1JJrFCbgO51cuXiy5Sx7SCSzEw-lWcVCV0oQ

Anthropic also recently published a blog on how they enable self-service analytics using LLMs and agent skills: https://claude.com/blog/how-anthropic-enables-self-service-data-analytics-with-claude.

This task is to investigate how these types of interface can support users of the data platform. For example, can the chat interface replace the need to teach users SQL for ad-hoc queries?

Review the Apache Superset extension API in pull request 41205 and Anthropic's self-service analytics blog first. Compare chat-based interfaces with the platform's current data-consumer workflows, especially ad-hoc queries and SQL training. Done means a documented investigation with findings, limitations, and a clearly scoped recommendation.

Written by the indexing model from the issue text.