P2.2+P2.3: NATS offload node-worker + router offload integration

Node Worker (services/node-worker/):
- NATS subscriber for node.{NODE_ID}.llm.request / vision.request
- Canonical JobRequest/JobResponse envelope (Pydantic)
- Idempotency cache (TTL 10min) with inflight dedup
- Deadline enforcement (DEADLINE_EXCEEDED on expired jobs)
- Concurrency limiter (semaphore, returns busy)
- Ollama + Swapper vision providers

Router offload (services/router/offload_client.py):
- NATS req/reply with configurable retries
- Circuit breaker per node+type (3 fails/60s → open 120s)
- Concurrency semaphore for remote requests

Model selection (services/router/model_select.py):
- exclude_nodes parameter for circuit-broken nodes
- force_local flag for fallback re-selection
- Integrated circuit breaker state awareness

Router /infer pipeline:
- Remote offload path when NCS selects remote node
- Automatic fallback: exclude failed node → force_local re-select
- Deadline propagation from router to node-worker

Tests: 17 unit tests (idempotency, deadline, circuit breaker)
Docs: ops/offload_routing.md (subjects, envelope, verification)
Made-with: Cursor

services/node-worker/models.py

@@ -0,0 +1,47 @@
+"""Canonical job envelope for cross-node inference offload."""
+from typing import Any, Dict, List, Literal, Optional
+from pydantic import BaseModel, Field
+import time
+import uuid
+
+
+def _ulid() -> str:
+    return str(uuid.uuid4())
+
+
+class JobRequest(BaseModel):
+    job_id: str = Field(default_factory=_ulid)
+    trace_id: str = ""
+    actor_agent_id: str = ""
+    target_agent_id: str = ""
+    required_type: Literal["llm", "vision", "stt", "tts"] = "llm"
+    deadline_ts: int = 0
+    idempotency_key: str = ""
+    payload: Dict[str, Any] = Field(default_factory=dict)
+    hints: Dict[str, Any] = Field(default_factory=dict)
+
+    def remaining_ms(self) -> int:
+        if self.deadline_ts <= 0:
+            return 30_000
+        return max(0, self.deadline_ts - int(time.time() * 1000))
+
+    def effective_idem_key(self) -> str:
+        return self.idempotency_key or self.job_id
+
+
+class JobError(BaseModel):
+    code: str = "UNKNOWN"
+    message: str = ""
+
+
+class JobResponse(BaseModel):
+    job_id: str = ""
+    trace_id: str = ""
+    node_id: str = ""
+    status: Literal["ok", "busy", "timeout", "error"] = "ok"
+    provider: str = ""
+    model: str = ""
+    latency_ms: int = 0
+    result: Optional[Dict[str, Any]] = None
+    error: Optional[JobError] = None
+    cached: bool = False