Lalinference o2+eos fix
There was a segmentation fault in lalinference_o2 when running with the EOS parameterizations. This is because of a small bug (basically, I call two "Destroy" functions twice), which I tracked down and fixed in this MR. @john-veitch should recognize this and can probably quickly sign off on it. Additionally, the review pointed out that the weights/abscissaes for the gaussian quadrature integrations were truncated and not rounded properly. I update them here.