GoodRunsLists: thread-safety fixes and code cleanup

The main change is to remove the thread-unsafe caching of the previous run and LB number in TGoodRunsList::HasRun[LumiBlock]. If the std::map lookup is really too slow we need to implement this properly.

Also modernize the code (e.g. rely on compiler generated copy/assignment) and enable the thread-checker. The latter also required a few extra annotations in LumiCalc.

Not sure who is responsible for this package these days. Tagging a few experts: @tadej @krumnack @ponyisi @akraszna