Signal simulation performance
Some of the test files take an extraordinarily long time to run -
random/tollN.kat each take ~40 min! This slows down testing, and is largely due to the signal simulation. Now that signal inputs at components are in the interferometer matrix, rather than the RHS vector, the signal simulation coupling coefficients depend on the output of the carrier simulation e.g. for radiation pressure. As there's no way currently to determine whether the carrier output at a node has changed, the full matrix gets recomputed for each point.
We should think about splitting up the various methods in
finesse/components/matrixfill.pyx, as many of the calculations are done needlessly e.g. radiation pressure couplings when a mirror has no mass set, recomputing p1->p2 etc. for all components in a signal simulation even when nothing's changed