F matrix convention inconsistencies #15
Description
Results seem to depend on what convention is used in add_red_noise. For now we reverted to using the same convention as libstempo (#14), but it would be good to better understand why this is an issue at all. In principle the change should only result in a phase shift, so while a given realization can be different, on average this shouldn't result in any systematic bias.