chore: fix transient random test failures

[paulbalandan]

Description
Random tests still fails at some points. Trying to probe the likely culprit.

Checklist:

• Securely signed commits
• Component(s) with PHPDoc blocks, only if necessary or adds value (without duplication)
• Unit testing, with >80% coverage
• User guide updated
• Conforms to style guide

[paulbalandan]

The fix here is to acknowledge the fact that the transient failures may be caused by race issues within GitHub Actions. Thus, the script will retry for another round with a fresh random seed. Given that the failures all revolve around CodeIgniter\CodeIgniter not found or failed to autoload, I disabled opcache in CLI as a primary defense and the retry as fallback.

The retry mechanism absorbed the flaky test fails in Language and Log in the SQLSRV matrix.

[michalsn]

I've tried various things myself, but since the problem only occurs on GitHub Actions, it's hard to debug.

The retry mechanism seems to work, but maybe we should stop running these tests in parallel?

[paulbalandan]

That could work as well. However, running the tests sequentially could take much more time as we incrementally add components, and the default repeat in GitHub Actions is 10.

Will that trade off be ok with you?

[michalsn]

Okay, let's check how this will work for us now.