feat: Provider model_settings 全链路 + 监管节点工具集 + 重型插件注入 + 前端打磨

- Provider model_settings (Provider+Model 级别参数配置): DB JSONB → API → GSM → AgentFactory.resolve → 三节点 agent.run 注入
- 新增 data/toolset/regulatory_toolset/: 监管节点专属工具（query_workflow_status / query_task_list / send_file）
- send_file 从 interactive_toolset 迁移至 regulatory_toolset，interactive 仅保留 approval
- mcp_helper 合入 GlobalPluginManager dispatch tools
- 前端 Provider 弹窗参数设置区加 JSON 编辑器（model_settings）
- 前端 Plugin 页面新增"重型插件"Tab（HeavyPluginList 占位）
- .gitignore 精简：去除系统默认项，修复 data/ 子目录追踪
- data/toolset/ 与 data/plugin/ 首次纳入版本控制

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

kilostar/core/individual/regulatory_node/regulatory_node.py

@@ -40,6 +40,7 @@ class RegulatoryNode:
 
         self.logger = get_logger("regulatory_node")
         self.agent: None | Agent = None
+        self._model_settings: dict = {}
 
     async def create_agent(
         self,
@@ -88,6 +89,7 @@ class RegulatoryNode:
             tools=tools,
             toolsets=toolsets,
         )
+        self._model_settings = AgentFactory.resolve_model_settings(provider, model_id)
 
         @self.agent.system_prompt
         async def dynamic_prompt(ctx: RunContext[RegulatoryNodeDeps]):
@@ -178,6 +180,7 @@ class RegulatoryNode:
                 instructions=self._CHAT_INSTRUCTIONS,
                 event_stream_handler=_stream_handler,
                 message_history=message_history,
+                model_settings=self._model_settings or None,
             )
         except Exception as e:
             self.logger.exception(f"RegulatoryNode.stream_working failed: {e}")
@@ -204,6 +207,7 @@ class RegulatoryNode:
                 user_prompt=message,
                 deps=deps,
                 message_history=message_history,
+                model_settings=self._model_settings or None,
             )
             response: MessageResponse = agent_response.output
             response.platform = platform