Reliablility report is inaccurate

I started investigating https://github.com/nodejs/node/issues/44805
when taking a look at https://github.com/nodejs/reliability/issues/388:

not ok 3789 sequential/test-watch-mode-inspect
  ---
  duration_ms: 120.67
  severity: fail
  exitcode: -15
  stack: |-
    timeout
    TAP version 13
    (node:37761) ExperimentalWarning: The test runner is an experimental feature. This feature could change at any time
    (Use `node --trace-warnings ...` to show where the warning was created)
  ...

the first PR in the Failed PR row has two CI runs prior to this report: 46819, 46808

the first CI run reports a failure for this based on the previous osx1015 run https://ci.nodejs.org/job/node-test-commit-osx/47578/ that did not actually run in this build (due to java.net.ConnectException: Connection timed out) but Jenkins still considers the child of a previous run to be the child of this run.

this obviously happened in other PRS/CI runs as well, wich makes the report links and count very hard to rely on when investigating flaky test issues

Start by tracing how the reliability report for nodejs/reliability#388 derives its Failed PR, First CI, Last CI, and Appeared links from the referenced Jenkins runs 46819, 46808, 47578, and 46857. Compare parent and child build relationships, including the failed connection in run 47578. Done means report links and counts reflect CI runs that actually executed.