test(training): verify 2v2 evaluator command

training/evaluate.py

@@ -82,6 +82,8 @@ def evaluate_pair(
     """Replay one seeded state sequence with the models on opposite sides."""
     if episodes < 2 or episodes % 2 != 0:
         raise ValueError("--episodes must be an even number of at least 2 for paired side swaps")
+    if team_size not in (1, 2):
+        raise ValueError("team_size must be 1 or 2")
 
     episodes_per_side = episodes // 2
     first = run_half(