[ML] Adjust the behavior of skip_result custom rules

[valeriy42]

Currently, if we define skip_result it avoids updating distribution quantiles. This can change the impact on the anomaly score, although the anomaly probability remains the same.

Ideally, we need to change the behavior such that for skip_result, the algorithm behaves "normally", except it avoids writing records and changing aggregates.

\cc @tveasey

[valeriy42]

Maybe we should introduce result.s_ShouldGetAnomaly or something like that similar to result.s_ShouldUpdateModel to control this behavior explicitly.

[droberts195]

I don't think we should introduce a third user-facing config option. Updating the model but not the quantiles just sounds wrong if the rule was set up to update the model but not write out any anomalies.