feat(engine): 实现行动点系统

- 扩展 AgentState 添加 action_points, max_action_points, last_action_tick 字段
- 新增 ActionFeedback 模型用于返回行动执行结果
- 创建 action_points.py 模块实现行动点消耗与恢复逻辑
- 行动消耗表: vote=1, trigger_skill=2, influence=2, comment/support/chaos=0
- 每 tick 恢复 1 点行动点（不超过 max）
- 行动点不足时拒绝执行并返回失败反馈
- 新增 7 个测试用例，全部 37 个测试通过

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

engine-python/tests/test_step.py

@@ -541,3 +541,136 @@ def test_skill_requires_prerequisite(client):
     assert skills["unity"]["requires"] == ["festival_boost"]
     # 初始状态下 unity 未解锁
     assert skills["unity"]["unlocked"] is False
+
+
+# ==================== 行动点系统测试 ====================
+
+def test_action_points_exists_in_agent(client):
+    """测试 agent 包含行动点字段"""
+    resp = client.post("/step", json={"events": []})
+    data = resp.json()
+
+    for agent_id, agent in data["world_state"]["agents"].items():
+        assert "action_points" in agent
+        assert "max_action_points" in agent
+        assert "last_action_tick" in agent
+        assert agent["action_points"] >= 0
+        assert agent["max_action_points"] >= 1
+
+
+def test_action_feedbacks_in_response(client):
+    """测试响应包含 action_feedbacks 字段"""
+    resp = client.post("/step", json={"events": []})
+    data = resp.json()
+
+    assert "action_feedbacks" in data
+    assert isinstance(data["action_feedbacks"], list)
+
+
+def test_vote_consumes_action_points(client):
+    """测试投票消耗行动点"""
+    # alice 投票（消耗 1 点）
+    events = [
+        {"type": "vote", "faction": "optimists", "text": "", "user": "alice", "ts": 1}
+    ]
+    resp = client.post("/step", json={"events": events})
+    data = resp.json()
+
+    # 验证行动反馈
+    assert len(data["action_feedbacks"]) == 1
+    feedback = data["action_feedbacks"][0]
+    assert feedback["success"] is True
+    assert feedback["user"] == "alice"
+    # 初始 3 点，消耗 1 点，恢复 1 点 = 3 点
+    # 但消耗在恢复之后，所以是 3 - 1 = 2
+    assert feedback["remaining_ap"] == 2
+
+
+def test_insufficient_action_points_rejected(client):
+    """测试行动点不足时拒绝执行"""
+    # 重置状态，设置 alice 行动点为 0，last_action_tick 为当前 tick
+    # 这样在下一个 tick 恢复后只有 1 点
+    state = get_default_state()
+    state.tick = 10
+    state.agents["alice"].action_points = 0
+    state.agents["alice"].last_action_tick = 10  # 刚行动过，不会恢复
+    save_state(state)
+
+    # 尝试投票（需要 1 点），但恢复后有 1 点
+    # 为了测试不足，我们需要让 last_action_tick = tick，这样不会恢复
+    # 但 regenerate 检查的是 tick - last_action_tick >= 1
+    # tick 会先 +1 变成 11，11 - 10 = 1 >= 1，所以会恢复
+
+    # 重新设计：设置 last_action_tick 为未来的 tick
+    state = get_default_state()
+    state.tick = 10
+    state.agents["alice"].action_points = 0
+    state.agents["alice"].last_action_tick = 11  # 未来 tick，不会恢复
+    save_state(state)
+
+    events = [
+        {"type": "vote", "faction": "optimists", "text": "", "user": "alice", "ts": 10}
+    ]
+    resp = client.post("/step", json={"events": events})
+    data = resp.json()
+
+    # 验证被拒绝
+    assert len(data["action_feedbacks"]) == 1
+    feedback = data["action_feedbacks"][0]
+    assert feedback["success"] is False
+    assert "insufficient" in feedback["reason"]
+
+
+def test_action_points_regenerate_per_tick(client):
+    """测试行动点每 tick 恢复"""
+    # 设置 alice 行动点为 1
+    state = get_default_state()
+    state.agents["alice"].action_points = 1
+    state.agents["alice"].last_action_tick = 0
+    save_state(state)
+
+    # 执行一个 tick（无事件）
+    resp = client.post("/step", json={"events": []})
+    data = resp.json()
+
+    # 行动点应该恢复到 2（1 + 1）
+    assert data["world_state"]["agents"]["alice"]["action_points"] == 2
+
+
+def test_action_points_not_exceed_max(client):
+    """测试行动点不超过最大值"""
+    # 设置 alice 行动点为 3（已满）
+    state = get_default_state()
+    state.agents["alice"].action_points = 3
+    state.agents["alice"].max_action_points = 3
+    save_state(state)
+
+    # 执行多个 tick
+    for _ in range(3):
+        client.post("/step", json={"events": []})
+
+    resp = client.post("/step", json={"events": []})
+    data = resp.json()
+
+    # 行动点不应超过 max
+    assert data["world_state"]["agents"]["alice"]["action_points"] <= 3
+
+
+def test_zero_cost_action_always_allowed(client):
+    """测试 0 消耗的行动始终允许"""
+    # 设置 alice 行动点为 0
+    state = get_default_state()
+    state.agents["alice"].action_points = 0
+    save_state(state)
+
+    # comment 类型消耗 0 点
+    events = [
+        {"type": "comment", "text": "hello", "user": "alice", "ts": 1}
+    ]
+    resp = client.post("/step", json={"events": events})
+    data = resp.json()
+
+    # 应该成功执行
+    assert len(data["action_feedbacks"]) == 1
+    feedback = data["action_feedbacks"][0]
+    assert feedback["success"] is True