Adding embeddings/RAG support to the Koog-based AI agent in Confetti

A walkthrough of adding semantic search to the Confetti Compose Multiplatform app using Koog's embeddings and RAG modules. The implementation embeds conference session data using Gemini, stores vectors in a persistent backend via Okio, and exposes a SearchSessionsTool to a Koog AIAgent so it can answer topic-style queries (e.g. 'what AI talks are on?') even when session titles use different wording. Two separate indexes (title-only and title+description) are maintained to improve ranking, and platform-specific cache directories are wired through Koin DI.