[CI] Windows mingw/ucrt - test-all timeout failures
Recently there was another test-all timeout failure on a Windows mingw or ucrt job. These are intermittent failures, and the last indicated that the test suite had passed.
ruby-loco runs tests differently, but it has also had intermittent failures, which were due to a ruby process that wasn't closing, leaving a locked file (STDERR mapped to a file).
Stop-Process -Name ruby -Force -ErrorAction SilentlyContinue (PowerShell), which in Ruby's CI could be added right after the test-all line.
Again, this is intermittent. I haven't checked to see if it can be repro'd using the test seed value from a failing job.