Include w/w differences in usage ping

Description

We send back a bunch of cumulative data with the usage ping, which we then try to tease out to determine week-over-week changes, but because the usage ping data isn't perfect, we end up with a lot of noise and invalid data.

e.g. We report total number of pipelines created. Let's also report how many pipelines were created in the last week.

Proposal

We should have the usage ping itself send differential data so we have more accuracy.