LongTraceRL: Learning Long-Context Reasoning from Search Agent Trajectories with Rubric Rewards - View it on GitHub