Fix test_exclude_detection_from_training: use days_back and backfill_days correctly

The test was failing because detection_period parameter was overriding
backfill_days, causing the detection period to span all 14 days of data
instead of just the last 7 days. This left no training data when
exclude_detection_period_from_training was enabled.

Fixed by:
- Removing training_period and detection_period parameters
- Setting days_back: 14 (scoring window includes both periods)
- Setting backfill_days: 7 (detection period is last 7 days)

This properly splits the data into:
- Training period: days 14-8 (7 days of normal data)
- Detection period: days 7-1 (7 days of anomalous data)

Test now passes locally with postgres.

Co-Authored-By: Yosef Arbiv <[email protected]>

Author: devin-ai-integration[bot]

integration_tests/tests/test_event_freshness_anomalies.py

@@ -146,9 +146,8 @@ def test_exclude_detection_from_training(test_id: str, dbt_project: DbtProject):
     test_args_without_exclusion = {
         "event_timestamp_column": EVENT_TIMESTAMP_COLUMN,
         "update_timestamp_column": UPDATE_TIMESTAMP_COLUMN,
-        "training_period": {"period": "day", "count": 7},
-        "detection_period": {"period": "day", "count": 7},
-        "backfill_days": 7,  # Explicit backfill_days for exclusion logic
+        "days_back": 14,  # Scoring window: 14 days to include both training and detection
+        "backfill_days": 7,  # Detection period: last 7 days (days 7-1 before test_started_at)
         "time_bucket": {
             "period": "day",
             "count": 1,