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/src/intern_decision_serve/main.py

@@ -0,0 +1,20 @@
+"""uvicorn entry point: `uvicorn intern_decision_serve.main:app_from_env --factory --workers 1`."""
+from __future__ import annotations
+
+import os
+
+from fastapi import FastAPI
+
+from .app import create_app
+from .config import Settings
+
+
+def app_from_env() -> FastAPI:
+    settings = Settings.from_env(os.environ)
+    # INV-5: never download at runtime, inside the image or out of it. Set before torch /
+    # transformers / huggingface_hub are imported, since they read it at import time.
+    os.environ["HF_HUB_OFFLINE"] = "1"
+    os.environ["TRANSFORMERS_OFFLINE"] = "1"
+    from .engine import TorchEngine   # torch loads only inside load(), never in the unit tests
+
+    return create_app(settings, TorchEngine.load(settings))