Fix SND attribute values lost or left stale across ETL runs

[labkey-martyp]

Rationale

Stop the SND ETL losing attribute values when the event data and attribute data steps' incremental windows drift apart, and make a blanked source attribute clear its stored value instead of leaving it stale. The event data step deleted and recreated each merged row's exp object, which discarded its attribute values, and only the attribute data step restores them; because that step computes its own incremental window, a row whose change landed in the seconds between the two steps computing their windows was cleared by the first and never seen by the second, and its values were gone permanently. The delete dates from the days before the attribute data step existed and has been redundant since that step began replacing values property by property.

The attribute data source view also filtered blank values out, so a cleared attribute never reached the ETL and its old value survived. That was masked by the same delete, which is why it surfaces alongside this fix.

The view change is a source-database ALTER VIEW and deploys separately from the module. The module change is safe to deploy first, since no null-valued rows arrive until the view changes.

Related Pull Requests

• Log EventDataIds through the SND event and attribute data ETL steps #985 added the per-step EventDataId logging that pinpointed the lost rows.

Changes

• The event data merge no longer deletes and recreates each row's exp object, so attribute values survive a merge that the attribute data step's window does not also cover.
• The attribute data source view passes blank values through as null-valued rows instead of filtering them out, so a cleared attribute reaches the ETL.
• The attribute data step removes the stored value for a null-valued source row; a lookup value that fails to resolve still leaves the old value in place.
• The diagnostic that reported which values a merge was about to clear is removed, since nothing is cleared now.