API for more easily running parallel chains

## 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.

Start by reading optimize_stepsize and the sample API in the Julia package, then compare the arguments needed for MCMCWrapper parallel chains with those currently accepted by optimize_stepsize. Check how initial conditions and sample_args flow into sample. Done means stepsize optimization works with n_chain, MCMCThreads, and parallel-chain initial conditions, with behavior for default initial values clearly defined.

Written by the indexing model from the issue text.