fix: 后端 API 修复

- deps.py: 依赖注入优化
- matches.py: 比赛列表查询修复
- predict.py: 预测接口参数校验修复
- orchestrator.py: 多 Agent 编排逻辑修复

src/llm/agents/orchestrator.py

@@ -95,10 +95,10 @@ class MultiPredictResult:
     pred_1x2: str | None
     subjective_confidence: float | None
     reasoning: str | None
+    context: str
     agent_outputs: list[dict]
     agent_weights: dict | None
     status: str = "success"
-    context: str
     latency_ms: int | None = None
     prompt_tokens: int | None = None
     completion_tokens: int | None = None