Idempotency bugs in add_reference_concurrently

[ghiculescu]

The adding of foreign keys in add_reference_concurrently is not idempotent. Instead of being safely skipped here, an error is raised.

The root cause is the merging in of validate: false here. That gets passed through to the check here. It means that if you use add_reference_concurrently to create a validated foreign key, and then call add_reference_concurrently again, it will always raise, because the matching foreign key won't be found.

I'm not sure what the right approach to fixing this is yet but I wanted to add some failing tests to demonstrate the issue.

[ghiculescu]

I think this is expected behaviour, so I added this as a passing test (though I guess it would be nice if it was consistent with add_reference_concurrently).

[fatkodima]

Thank you for the PR!

[ghiculescu]

Thanks for the fix!

[ghiculescu]

We should have included this test too:

    def test_add_reference_concurrently_with_unvalidated_foreign_key_is_idempotent
      assert_empty connection.foreign_keys(:milestones)

      connection.add_reference_concurrently :milestones, :project, foreign_key: { validate: false }
      connection.add_reference_concurrently :milestones, :project, foreign_key: { validate: false }

      assert_equal 1, connection.foreign_keys(:milestones).size
    end

It passes with this fix, so all good.