Filtering the ruptures may become terribly slow

[Michele Simionato]

We discovered that when the parameter rupture_mesh_spacing becomes small (say 0.01) some tests (say qa_tests/hazard/event_based/case_4) becomes terribly slow (4 hours instead of 40 seconds).. In general, filtering the ruptures made the event_based calculator slower. This is an investigation ticket, to remind ourselves of the issue and see if it is possible to do something better.