run_all.sh reproduces every file under analysis/data/ from zero: it
clones the in-scope components if missing, exports group issue metadata
from git.lab, and regenerates the inventories. Reruns are diff-clean --
no wall-clock time enters an output; 'days since' is measured against
the management repo's HEAD date.
The management repo is inventoried at main, not at the analysis branch,
so this analysis does not observe its own commits. inv_repo.py aborts
the run if anything outside analysis/ was modified.
