Perform checks for anomalous values

[rich-iannone]

This PR introduces the col_anomaly_check() function. The goal is to catch anomalous values and signal the number of these anomalies as failing test units. This is also meant to work with all common database backends by using a combination of modeling on the user system and applying model predictions to the remote table via joins.

To be done:

• provide rows with anomalies as downloadable CSV in agent report
• generate ggplot artifacts that show how anomalies are detected (needs further feature for provision of supplemental materials)
• add more time granularities for time-window size aside from those that are currently offered
• deal with gaps in time-series data and window sizes that have low amounts of data points
• allow x data that is not date-time data
• allow for univariate anomaly detection
• allow user to choose a formula for mgcv::gam()

Fixes: #246