Factored transfer selection of binarization schemes: code, protocol, manifests, tables and per-run results

Everything needed to rerun the study of the same name, or to recompute every number of its manuscript from the runs already made. The record holds the study code, including the reference implementation the patch is applied to; the protocol, with its two dated amendments; the exact command line of every campaign; the policies trained on the training set, one per family and one per family and optimizer; one csv per instance, optimizer and scheme, with one row per seed; the per-generation traces of the convergence sample; and the outputs of the analysis scripts, which are what the tables and figures of the manuscript are drawn from. Nine campaigns are included. Three reference factorials of 34 binarization schemes were produced for this study, on the railway level of set covering, on feature selection and on unit commitment; the other four families reuse the factorial of the earlier calibrated layer study, which is deposited separately and cited in the manuscript. The remaining campaigns are the calibration phase, the main evaluation over 115 instances of six families, six optimizers and 31 seeds, the extreme-scale study on two railway instances of about one million columns, the cross-optimizer transfer study, an ablation with the paid oracle disabled, and a sample of runs with the per-generation trace switched on. Recomputing the numbers of the manuscript from the runs already here takes minutes; rerunning a campaign takes days. README.md gives both routes, and SHA256SUMS.txt a checksum for every file.