Feature: add automatic anomaly detection for column stats

[MiConnell]

Currently, column statistics are stored based on each run (row_count, sum, avg, etc). but the only way to tell if a large change has occurred is visually through a BI tool or manually with a query. Flags with a threshold would be very useful.

Something like the below variable

"dbt_observability.column_anonamly_threshold_percent": "5"

which would then flag any time that a column stat changed by more than 5%.

[ataft]

Perhaps this could be done similar to #28, but with a test on the column rather than the model