Does two-level write transforms.csv and accept the long CSV, and do the twolevel_* post-processing scripts delegate to the shared scripts?

[gdevenyi]

Part of #138

Question

Two-level fits the transform manifest decided in #144 (level 1: scan → subject Template; level 2: subject Template → population Template). Do we port it, and how far, while keeping its CLI, input format, and output layout backward compatible?

Proposed answer (from the #144 session):

1. Additive: twolevel_modelbuild.sh also writes transforms.csv and also accepts the long CSV (grouped by subject_id); nothing existing changes.
2. Validate the shared scripts by running longitudinal_dbm.sh and the shared resample scripts on two-level output and comparing with twolevel_dbm.sh and twolevel_*_resample.sh.
3. Later, optionally: the twolevel_* post-processing scripts become thin shims over the shared scripts (or are deprecated), only once step 2 shows identical outputs.