clean nhsn data

[aysim319]

related to #2152

The pipeline has an unexpected behavior where the update timestamp was changed and the sad path was to log the data as stale, but continue to process and update the data.

The run from 2025-06-25 had more unexpected behavior where the preliminary dataset that was recently updated still used the old key rowsUpdatedAt while the regular dataset did switch over to using viewLastModified which logged the data as stale, but ended up still processing the data

[melange396]

i think this should sorta fix itself on friday... when the pipeline runs on that day, the bad data that was erroneously/accidentally written today (from the "regular" dataset with last week's values but this week's issue date) will hopefully be overwritten with the correct data values (that have yet to be published). the good data written today (from the "prelim" dataset) should be untouched by that.

we should pay close attention to all that, though. the pipeline job is scheduled to run at 1pm, and afaict, the data seems like it typically gets published around noonish.