Rationalize logging in QtTestLib

At present the APIs for logging in QTest are a bit incoherent:

• the QTestLog methods addIncident() and addMessage() split things to be reported a bit eccentrically – B?XFAIL is never the decisive result of a test, while SKIP is, but the former is reported via addIncident() with the things that are decisive, while the latter is reported by addMessage(), along with others that aren't (although QCritical and QFatal typically accompany a FAIL that is); and addIncident() lacks a parameter from the caller that could tell it when the incident isn't decisive (as happens if the decisive incident didn't return from the test, e.g. a B?XPASS/Continue or a SKIP or B?FAIL in a helper function, such as an event handler in a test that uses an event loop).
• in addition, addMessage() rolls together internal logging from QTest, such as the Info output from -v1 and -v2 command-line options, with messages originating in the usual Qt logging infrastructure; while the enum does provide a way to separate these, it might make more sense to have separate functions, to separate test-framework messages from messages originating in the actual test or the code-under-test.
• the counting is out of whack and It's all undocumented OK, I've just fixed that ;^)

We need to do a design review and at least come up with a plan of action for how to move forward. vestbo has begun doing that (and I've helped a bit); this ticket is really a post hoc way to track our progress with work already begun ! It can also serve as a place to document what we decide and why, along with work we do to move forward with it.