Modify algorithm to specify conjugate priors for covariance matrices

[landmanbester]

The current form of the algorithm requires quite a bit of fine tuning to get things right i.e. to maintain over dispersed priors for all the inputs. This is especially true for the value of the cosmological constant when redshift drift data are incorporated. One way to fix this is to use (possibly scaled and/or separated see here inverse Wishart distributions to specify conjugate priors over the covariance matrices. As a start we need to, at the very least, specify a conjugate prior for the uncertainty in the cosmological constant since the variance in Lambda resulting in good acceptance rates in the parameter that requires most fine tuning.