How can we disable reasoning tokens for Openrouter models to reduce latency? Openrouter provides the option (screenshot), but can we set it through Vapi? https://cdn.discordapp.com/attachments/1211484481848873020/1402979859562627153/Screenshot_2025-08-07_at_12.40.39.png?ex=6895e250&is=689490d0&hm=c8524c051ad3f6a9c19302a0b2006aab247afc928cc86e72f927f728bd142a79&