feat(drive-and-save): add llm config block

Append llm: block to config.yaml with zerogpu_model_id, hf_model_id,
gpu_seconds, recommendation_max_sentences, and medical_disclaimer.
Add test_config_has_llm_block to verify the block is present and valid.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

config.yaml

@@ -27,3 +27,10 @@ funnel:
 ui:
   approx_label: true
   disclaimer: "Published negotiated rates, not quotes — your out-of-pocket depends on your plan."
+
+llm:
+  zerogpu_model_id: "microsoft/Phi-4-mini-instruct"   # env ZEROGPU_MODEL_ID overrides
+  hf_model_id: "microsoft/Phi-4-mini-instruct"        # serverless fallback (needs an Inference Provider; local-dev only)
+  gpu_seconds: 45                                      # env ZEROGPU_DURATION_SECONDS overrides (env wins)
+  recommendation_max_sentences: 3
+  medical_disclaimer: "This compares published negotiated prices, not your out-of-pocket cost, and is not medical advice."

test_core.py

@@ -230,3 +230,14 @@ def test_build_demo_returns_blocks():
     import gradio as gr
     demo = app_module.build_demo()
     assert isinstance(demo, gr.Blocks)
+
+
+# ---------------------------------------------------------------------------
+# Part I — llm config + module
+# ---------------------------------------------------------------------------
+
+def test_config_has_llm_block():
+    cfg = core.load_config()
+    assert cfg["llm"]["zerogpu_model_id"]
+    assert cfg["llm"]["recommendation_max_sentences"] >= 1
+    assert "not medical advice" in cfg["llm"]["medical_disclaimer"]