Multichain for rubias

[boppingshoe]

• To run multiple chain, set nchains = (pick a number) in run_rubias_mix(). For single chain, nchain = 1 (default)
• Output from run_rubias_mix() saved as .csv files have a new column "chain"
• Outputs created previously can still be summarized by the updated custom_comb_rubias_output() and stratified_estimator_rubias() (backward compatible)

[awbarclay]

I tested run_rubias_mix and custom_combine_rubias with 7 mixtures, including some that have known convergence issues. With the full model "BR", I got very similar Gelman-Rubin shrink factors as I did previously with BAYES. Running without the full model "MCMC" also worked, but as expected, all chains converged. custom_combine_rubias summarizes the output object from run_rubias_mix and the output files it produces. However, when I try to summarize old output files without the "chain" variable, I get the following error message:

Summarizing results from rubias output .csv files.
    Building trace output from `repunit_trace.csv` files.
Error in { : task 1 failed - "Can't subset columns that don't exist.
✖ Column `chain` doesn't exist."

Looks like custome_combine_rubias needs some work to make it backwards compatible with the old output files.

I have not tested stratified_estimator_rubias yet.

[boppingshoe]

@awbarclay the issue with older files is corrected. I think