refactor tests that depend(ed) on the Config enum being functional

some tests explicitely call on members of the Config enum to perform tests. refactor those tests so the enum can be phased out itself.

> 2. Tests that mock ConfigurationDao.getValue(...) directly, where switching the production code to .value() would make the mock inert (since .value() reads from the static ConfigDepot, not the injected DAO):
> - ApiRateLimitServiceImpl.java's ApiLimitEnabled/ApiLimitInterval/ApiLimitMax/ApiLimitCacheSize reads — ApiRateLimitTest.java mocks these with a value ("true" for enabled) that diverges from the ConfigKey defaults, so skipping here isn't optional.
> - FirstFitPlanner.java's ImplicitHostTags read — FirstFitPlannerTest.java mocks it to "GPU", which also diverges from default.
> - ScaleIOPrimaryDataStoreDriver.java's MigrateWait read — deferred out of caution: ScaleIOPrimaryDataStoreDriverTest.java mocks it, and since ConfigKey's depot state is a static field, there's cross-test-pollution risk from other test classes (e.g. FirstFitPlannerTest) calling ConfigKey.init() in the same Surefire fork — not something to resolve with a coincidental "the mocked value happens to match the default" argument.
> - (ConfigurationManagerTest.java's ControlCidr and ActionEventUtilsTest/ActionEventInterceptorTest's PublishActionEvent were also mocked directly, but their mocked values matched the ConfigKey defaults exactly, so those were judged safe to migrate rather than deferred.)