SWE-EVO: Benchmarking Coding Agents in Long-Horizon Software Evolution Scenarios | Read Paper on Bytez