Refactor environment and per-rollout tool setup in GRPOTrainer , DPPOTrainer , GRPOWithReplayBufferTrainer …

Refactor environment and per-rollout tool setup in GRPOTrainer , DPPOTrainer , GRPOWithReplayBufferTrainer …: a task in HF ML Tasksmith (Harbor dataset). At present, providing environment factory eagerly constructs a fixed pool at initialization and builds tool dictionaries by slot. Replace that…