feat(intern-decision-serve): Intern-Decision-4B behind semif-serve's HTTP surface

Contract, service and tests (fake engine, no GPU). Scores through the checkpoint's own
inference.py (DecisionEngine.predict, sha256-pinned); maps semif decisions onto Jev choice
questions, packs /decide/shared into calls of at most 16, runs orderings in waves, and keeps
semif's error mapping, admission, body limit and hard VRAM cap. Deltas from semif-serve are
listed in the contract.

services/intern-decision-serve/tests/test_main.py

@@ -0,0 +1,23 @@
+"""Entry point: INV-5, offline before torch/transformers load. Contract: intern-decision-serve.contract.md."""
+import os
+
+from fake_engine import FakeEngine
+
+
+def test_app_from_env_goes_offline_before_the_engine_loads(monkeypatch):
+    from intern_decision_serve import engine as engine_module
+    from intern_decision_serve import main
+    seen = {}
+
+    def fake_load(settings, **_kw):
+        seen["offline"] = (os.environ.get("HF_HUB_OFFLINE"), os.environ.get("TRANSFORMERS_OFFLINE"))
+        seen["token"] = settings.api_token
+        return FakeEngine()
+
+    monkeypatch.delenv("HF_HUB_OFFLINE", raising=False)
+    monkeypatch.delenv("TRANSFORMERS_OFFLINE", raising=False)
+    monkeypatch.setenv("INTERN_DECISION_API_TOKEN", "k" * 40)
+    monkeypatch.setattr(engine_module.TorchEngine, "load", staticmethod(fake_load))
+    app = main.app_from_env()
+    assert seen == {"offline": ("1", "1"), "token": "k" * 40}
+    assert app.title == "intern-decision-serve"