fix DSML token leak and use no-thinking model for all non-tool-bound paths

add deepseek compatibility

lint

feat: apply gemini cooldown to async model calls

feat: apply cooldown to bound model calls

feat: track batch rate limit window

feat: track per-question rate limit streaks

feat: abort batch on gemini daily quota

feat: add gemini cooldown registry

feat: classify gemini client rate limits

feat: implement web search and file processing tools with robust retrieval fallbacks