feat(agent): unify MCP tools and adopt Responses delivery

Move project and host queries into native MCP with host-owned OAuth/context,
while retaining memory, skills, actions and publication plugins. Add nearby-model
queries and explicit chat/task/learning model defaults.

Adopt Responses HTTP/SSE and native steer/queue inbox delivery, preserving scoped
session ownership, pending input edits and investigation communication.

Disable local-provider discovery: Server /health allowed native vLLM to probe
/v1/models on port 8000. A native before/after regression prevents recurrence.

Validation: 170 focused Bun tests; service/plugin typechecks; native discovery,
provider-discovery and input-inbox smoke checks; no paid model inference.

agent-models.json

@@ -1,7 +1,9 @@
 {
-  "default_model": "waternetwork/gpt-6-luna",
-  "task_model": "waternetwork/gpt-6-luna",
-  "learning_gate_model": "waternetwork/gpt-6-luna",
+  "defaults": {
+    "chat": "waternetwork/gpt-6-luna",
+    "task": "waternetwork/gpt-6-luna",
+    "learning_gate": "waternetwork/gpt-6-luna"
+  },
   "providers": {
     "waternetwork": {
       "name": "WaterNetwork",
@@ -21,7 +23,7 @@
           "description": "深入分析和复杂任务处理",
           "icon": "sparkle",
           "limit": {
-            "context": 272000
+            "context": 475000
           }
         }
       }