Fix: take CI results into account in the "last updated" and "total time in review" columns

Right now, the algorithm for computing a PR's state over time only looks at label and draft status changes, not at the CI status. Fixing this is easy in principle (I have a branch doing so), but runs into a tricky bug:

There is no good easy way to determine when a commit was pushed to a pull request.

A commit has an authored date and commit date (which could be later if e.g. a commit was rebased) --- but we care about the time of the push. Github's graphql doesn't expose this information [any more] (and the pushDate field also was not reliable before).

One could query the REST API for github pushes to a pull request: that seems like it should work, but requiring additional API calls is not nice. We'd like to avoid doing this all over; some engineering could certainly help

• only ask for push dates when CI actually fails for the first time, and passes again
• cache the results somehow. Probably, it's useful to have a separate JSON file, and the analysis merging the information in these every time. Implementable, but not fun!

Another option: live with values that are sometimes wrong.
Is there a third, better way I don't know yet?

Start by tracing the algorithm that computes a pull request's state over time, focusing on how label and draft-status changes are currently used. Compare the available commit dates with GitHub GraphQL and REST push data, then define a reliable approach that accounts for CI status and produces correct “last updated” and “total time in review” values.