This PR addresses two different audiences:

• The side that covers users of Slint is Rust specific (cargo etc.). I think that should be generalised and perhaps could be a PR of its own to be reviewed.
• The side that covers developers working on Slint appears to go into odd levels of details (see binary search for cubic bezier animation curves) - it feels like a "cache" where the AI summarised its findings and would use that for subsequent queries without ingesting the same over and over again. Olivier suggests to place these kind of "cache" documents into a dedicated folder that makes it clear that the content itself is AI generated, for use at your own risk. The top-level docs/ folder then stays clean for documentation that's intended for humans first.