Stop running known failures for Windows tests
Overview
We have some known failures on our Windows tests that we are trying to solve in &888. At the moment the known failures are allowed to fail by using the test failures. We then have a jobs that check if these are new failures or known failures.
We end up wasting time and compute running tests that we know are going to fail.
Proposal
A better approach would be to entirely skip running the tests that we know that are going to fail so that we save up time and compute. This might go hand in hand with #4467 (closed) since we'll be able to remove the tests that we know are going to fail