Flatten the list of rescaled values in a ConditionalPriorDict

I was recently contacted about an error that occurs if trying to use a MultivariateGaussian prior for a BBH analysis, where the prior is initially set up as a BBHPriorDict and the "ra" and "dec" keys are then altered to be MultivariateGaussian priors. The prior dictionary is therefore based on a ConditionalPriorDict, which has a different rescale method than a standard PriorDict. The rescale method of the standard PriorDict flattens the returned list to take account of the MultivariateGaussian's returning all its parameters in a single array/list rather than separately. However, this flattening was not in the ConditionalPriorDict's rescale method, so this is now added.

This allows joint priors on parameters that are within a ConditionalPriorDict.