Harnessing multiple cores to speed up fits with small number of chains

[ckrapu]

To add on to the previous question, are you interested in the parameter space position of the tuned chain, the learned covariance matrix, or both?