those were leftovers of bad reshuffling done earlier.

let's just keep the upgrades stuff in the upgrades
section. also. metrics was moved to monitoring.

The linear regression stuff makes zero sense in the long term anyways.

This record is adding noise to the stacked graph I'm developing, and
the data is already present in the dataset, in a way. Specifically, 5
days later, we have a proper record for all release types, where
jessie is zero. By missing the other releases, we make it look like
all all other releases are also zero, implicitly, so this is really
just corrupted data.