WIP: reduce number of calls to gps_time_to_gmst
Rather than look for a speed up in the conversion between gps time and gmst this branch just reduces the number of evaluations to one per likelihood evaluation.
This has sped up the likelihood by more than 6x for the three interferometer case.
I don't think this is an optimal solution and if someone else can write a conversion function that runs is much less than 1ms then I'll happily ditch this branch.