daily_row_counts (table_name, load_date, row_count) records how many rows each table received per day. A day with no load has no row.
Flag a table's day as an anomaly when its row_count is more than 50% above or below its baseline: the average row_count of the same table over the 7 calendar days before it (load_date − 7 through load_date − 1 — the day itself is not part of its own baseline). Only days that have a row count towards the baseline, and a day is only judged when its baseline comes from at least 3 such days; otherwise it's skipped.
Columns: table_name, load_date, row_count, baseline (rounded to 1 decimal place), pct_change (the difference from the baseline as a percentage of the baseline, signed, rounded to 1 decimal place). Return only the anomalies, sorted by table_name, then load_date.