feat: database setup with SQLAlchemy ORM models for Device and DeviceLog

- Replace minimal database.py with full SQLAlchemy async setup
- Create Device and DeviceLog models with proper relationships
- Add tests for model creation and relationship functionality
- Fixed lazy loading issue in tests by avoiding .count() on dynamic relationships

tests/test_models.py

@@ -0,0 +1,49 @@
+import pytest
+from mqtt_home.models import Device, DeviceLog
+
+
+async def test_create_device(db_session):
+    device = Device(
+        name="客厅灯",
+        type="light",
+        protocol="custom",
+        mqtt_topic="home/living/light",
+        command_topic="home/living/light/set",
+    )
+    db_session.add(device)
+    await db_session.commit()
+    await db_session.refresh(device)
+
+    assert device.id is not None
+    assert len(device.id) == 36
+    assert device.name == "客厅灯"
+    assert device.state is None
+    assert device.is_online is False
+
+
+async def test_create_device_log(db_session):
+    device = Device(
+        name="温度传感器",
+        type="sensor",
+        protocol="custom",
+        mqtt_topic="home/temperature",
+    )
+    db_session.add(device)
+    await db_session.flush()
+
+    log = DeviceLog(
+        device_id=device.id,
+        direction="rx",
+        topic="home/temperature",
+        payload='{"temperature": 25.5}',
+    )
+    db_session.add(log)
+    await db_session.commit()
+
+    # Query logs directly instead of using relationship.count()
+    from sqlalchemy import select
+    result = await db_session.execute(
+        select(DeviceLog).where(DeviceLog.device_id == device.id)
+    )
+    logs = result.scalars().all()
+    assert len(logs) == 1