Stop disabling external schemes via QDesktopServices in tests

Some tests use a bogus protocol to trigger an error condition (ie, failed load). Since bug 1384460 landed, these bogus protocols get handed off to QDesktopServices which results in the load being cancelled. To work around this, there is a test API to tell QDesktopServices to not handle the bogus protocol, restoring the old behaviour (load failing).

Rather than doing this we should adapt the tests to accept the updated beahviour (which I assume is that we get a stopped load rather than a failed load), as this is closer to the real-world behaviour. Tests that need a failed load should create it another way (eg, using an invalid http URL)