[regression] Flag global property should not be changed on existing flags

This has caused me to dig up this very old issue: #882564: Add assertion helpers to AssertContentTrait for disabled/enabled fields.

Oh dear, the tests are doing the opposite of what they're meant to!

I had the assertion the wrong way round...

Drat! I didn't actually read the test result, as I only had about 5 minutes to reroll!

> Move to a new test that creates a global flag and doesn't do the part about enabling and disabling the flag again?

Yup, that makes sense.

I was working on this one too...

I managed to get my refactoring to pass last night, but didn't have time to clean up and post a patch.

+++ b/src/Tests/FlagTestBase.php
@@ -68,11 +68,13 @@ abstract class FlagTestBase extends WebTestBase {
-  protected function createFlag($entity_type = 'node', $bundles = [], $link_type = 'reload') {
+  protected function createFlag($entity_type = 'node', $bundles = [], $link_type = 'reload', $global = FALSE) {

I pondered doing this, but I feel the more parameters we add to this helper method, the less helpful it becomes -- at some point, it will simply have parameters for all of the flag properties, and then it might as well have a single $config param, and then we're no longer saving very much effort at all.

My take on refactoring the tests is to move both of doFlagLinksTest/doFlagGlobalLinksTest to a new test that focuses on access to a flag based on ownership. In due course we can add a test class for access based on permission, and also the access_author option.

Committed.