feat(ai-services): add AVM v2 A/B comparison endpoint and tests

Add POST /avm/v2/compare-v1 endpoint that runs both v1 (single-model)
and v2 (ensemble) AVM predictions on the same property and returns a
side-by-side comparison with price diff, confidence delta, and a
recommendation on which model to prefer.

- ABComparisonRequest/Response schemas in avm_v2 models
- compare_v1() method in AVMv2EnsembleService
- 4 new integration tests for the comparison endpoint
- All 47 Python tests pass

Co-Authored-By: Paperclip <noreply@paperclip.ing>

libs/ai-services/tests/test_avm_v2.py

@@ -172,3 +172,75 @@ def test_model_info_v2():
     data = resp.json()
     assert "model_version" in data
     assert data["is_active"] is True
+
+
+# ── A/B comparison tests ─────────────────────────────────────
+
+_COMPARE_PAYLOAD = {
+    "district": "Cầu Giấy",
+    "city": "Hà Nội",
+    "property_type": "apartment",
+    "area_m2": 80.0,
+    "rooms": 2,
+    "month": 3,
+    "quarter": 1,
+}
+
+
+def test_compare_v1_returns_both_models():
+    """Compare endpoint returns v1 and v2 predictions."""
+    resp = client.post("/avm/v2/compare-v1", json=_COMPARE_PAYLOAD)
+    assert resp.status_code == 200
+    data = resp.json()
+
+    assert "v1" in data
+    assert "v2" in data
+    assert data["v1"]["estimated_price_vnd"] > 0
+    assert data["v2"]["estimated_price_vnd"] > 0
+    assert 0 <= data["v1"]["confidence"] <= 1
+    assert 0 <= data["v2"]["confidence"] <= 1
+
+
+def test_compare_v1_returns_diffs():
+    """Compare endpoint computes price and confidence differences."""
+    resp = client.post("/avm/v2/compare-v1", json=_COMPARE_PAYLOAD)
+    data = resp.json()
+
+    expected_diff = data["v2"]["estimated_price_vnd"] - data["v1"]["estimated_price_vnd"]
+    assert abs(data["price_diff_vnd"] - expected_diff) < 10_000  # rounding tolerance
+
+    assert "price_diff_pct" in data
+    assert isinstance(data["price_diff_pct"], float)
+    assert "confidence_diff" in data
+
+
+def test_compare_v1_returns_recommendation():
+    """Compare endpoint provides a recommendation string."""
+    resp = client.post("/avm/v2/compare-v1", json=_COMPARE_PAYLOAD)
+    data = resp.json()
+
+    assert "recommendation" in data
+    assert len(data["recommendation"]) > 0
+
+
+def test_compare_v1_with_v2_features():
+    """Compare endpoint passes v2-specific features correctly."""
+    payload = {
+        **_COMPARE_PAYLOAD,
+        "distance_to_cbd_km": 5.0,
+        "distance_to_metro_km": 0.8,
+        "flood_zone_risk": 0.1,
+        "building_age_years": 3,
+        "has_elevator": True,
+        "has_parking": True,
+        "renovation_score": 0.9,
+        "view_quality": 0.8,
+        "interior_quality": 0.85,
+    }
+    resp = client.post("/avm/v2/compare-v1", json=payload)
+    assert resp.status_code == 200
+    data = resp.json()
+
+    # v2 should capture these extra features
+    assert data["v2"]["estimated_price_vnd"] > 0
+    assert data["v2"]["model_version"] is not None