llama-server's Responses API marks truncated replies and tool calls as completed

TL;DR: llama-server's OpenAI-compatible /v1/responses endpoint never reports a reply as cut off. On...