A surplus production model (SPM) implemented using RTMB. Computes the
negative log-likelihood (NLL) from observed and predicted CPUE data. The
scalar cpue_creep parameter supplies the proportional linear increase in
catchability per observation within each CPUE-series group.