Limit running of Windows Test
Overview
The Windows test take a very long time to complete, and we should only run them when they are really needed. In slack there was this "Also starting two full pipelines (with windows jobs!) for a single character typo fix seemed to be a huge overhead" and simply becuase they are heavy tests and we don't have a lot of windows machines (only 1 per version)
Proposal
Limit when to run the windows test:
- Run them only for merge requests.
- Run them only when we change a go file, any windows Dockerfile