tst_QDateTime::daylightTransitions() makes invalid assumptions

It assumes particular behaviour for the handling of times in the gap of a spring-forward or the fold of a fall-back. The resolution of such times is essentially arbitrary, although we have some reasonable heuristics for the results of adding an offset from a time before or after, if it knew its offset. I need to re-write this test slowly; we need to blacklist it in the mean time.