[test] Karma should fail if errors are thrown

[oliviertassinari]

I have based the pull request on top of the two wrong commits that were reverted. The expected outcome is that the CI to be failing

A post-mortem

• Everything starts with [DataGrid] Add fluid columns width support #480. The pull request is depending on a week old commit. Damien pushed some refactorization around the state management that [DataGrid] Add fluid columns width support #480 didn't include when running the tests (based on a week old commit). Once we merged, the changes conflicts and broke one of the tests in the CI (on master).
• As a forward-looking pragmatic approach, we started looking into how we could fix the issue (instead of reverting). Damien made a pull request Fix columns #542, probably based on a gut feeling that the logic should be written in a different way.
  Damien got a positive feedback loop: the tests are now green!
• 💥 Looking closer, it turns out that. The feedback loop was plain wrong. The new logic was introducing an error (setTimeout) that was preventing all the tests in the CI to run. Basically, only the first test is run, instead of the whole suite.
• I revert the last two commits, we have made things worse. Time to hit the reset button.

Before

After exit code: 0 🙀

Learnings

• Danail: Rebasing on master is a healthy practice :). When large changes are done, and a pull request starts to be old, best to rebase it. (I personally prefer to rebase much more than merge because I find resolving conflicts a lot easier) [DataGrid] Add fluid columns width support #480.
• Damien: We should test our changes a bit more :). Starting the documentation or the storybook or the failing test case when making Fix columns #542 would have prevented the issue in the first place.
• Olivier: The CI should have failed, found Uncaught exceptions in between Mocha tests are not reported karma-runner/karma-mocha#227, applied it here. It will prevent this issue to happen again.

Next step

• Apply [DataGrid] Add fluid columns width support #480 back :) cc @DanailH
• Apply the fix in the main repo.