Proposal: Track flaky tests as bugs

[flaneur2020]

Background

the last 0.7.126-nightly build did not actually pass the sqlogictest.

currently we did not forced all test to pass before merging, this is mostly due to there're some flaky tests exists, we still can merge the Pull Request as we found some test failures are not caused by my pull request.

flaky tests harms our confidence about automated tests, if the wolf really comes (like this time), we'd likely to dis-trust the alert, and causes real broken build.

Proposal

maybe we can deal with the flaky test as the following procedure:

1. found a flaky test, confirm it's not caused by my pull request
2. raises an issue about flaky test, mark the test as flaky (so it could be skipped in the Pull Request CI jobs), and reference the issue's link in the code comment
3. the reviewer confirms the skipped flaky test is tracked in issue, other wise do not LGTM at all.

so we can force each pull request MUST pass all the tests before merging, and build a trust with the CI.