Add FSP multi-agent architecture: Red Team LLM action space + alternating turns

Initial commit of CyberSOC upgraded RLVR environment