API for more easily running parallel chains

[odunbar]

Issue

Currently one can run sample with the MCMCWrapper object with n_chain parallel chains, by

• providing a n_chain-length vector of ICs (one per chain)
• providing the n_chain and MCMCThreads etc. to sample(..)
  But the stepsize calculation optimize_stepsize doesnt yet support the additional args.

Possible solution

It may be that one can quickly provide extension to optimize_stepsize to accept additional sample_args that can be internally passed into sample and this will resolve all issues. One could also allow for the default initial chain values to be extended to parallel chains by creating a vector of ICs equal to the prior mean.