Complex Fault Refactoring

[mmpagani]

In this PR we are replacing the original sampling method with the the new one implemented in the line module. This new approach introduces slight differences in the geometries of the surfaces generated but - overall - the results appear consistent with the original ones.

Note that several QA tests pass locally but require increased tolerance when executed on the remote machine. The disaggregation QA test 7 shows slightly larger differences between the hazard computed with the previous and current sampling methods. This depends on the fact that the site is on one side of the two subduction sources used and the new sampling algorithm creates a slightly shorter mesh. See the figure below where the blue dots show the new mesh and the pink dots are the old one. The point in yellow is the site.