Add trace button for thinking actions

Similar to how tool calls get a trace button, thinking actions should also get a clickable trace button so the user can see what the agent was thinking. This improves transparency and debuggability of agent reasoning.