Analyze log ouputs of failed tests, especially those that fail in a column unexpectedly

[antter]

Is your feature request related to a problem? Please describe.
As noted in #447 , our infra flake detection is built on the wrong assumptions and not really a good model. In #448 I redefined an infra flake as a bundle of tests that fail unexpectedly, and made a definition for this "unexpectedness." Tests can fail simultaneously for several reasons, an infra flake being only one example.

Describe the solution you'd like
A notebook that explores the log outputs of run tests, especially the tests that happen in an "unexpected" manner.

[antter]

First step that I am working on: collecting the data in a parallel manner.