As a DAG engineer,
I'd like to evict change points before data at the tip of a metric's time series, and recalculate the change points in order to reflect the latest test data.
Such that, the change points are kept up to date.
- Hook into Cedar's event infrastructure and raise a "new data at tip of metric XYZ" event for each statistic in a rollup when a new test run event is raised
- Create handler that handles these events by evicting all change points 50-10,000 (configurable) points before the tip of the time series, recalculates them, and persists them (in one transaction)
- Handler should exit gracefully, notifying cedar of success or errors to look into