Recursively walks a directory, peeks each file's 4-byte magic to
identify the format, derives the per-format --validate-X flag
from the format table's --info-X entry, and invokes that
validator as a subprocess for each file. Reports total / passed
/ failed / skipped counts and lists the failure paths with their
exit codes; returns exit 1 if any failure is found.
For each failure it prints the exact follow-up command needed to
reproduce the detailed error message, so the user doesn't have
to remember which validator goes with which extension. Asset-
style formats with no validator (.wom / .wob / .whm world
geometry) are counted in the "skipped" bucket but don't fail
the run.
Composes with the existing audit/fix tooling:
--audit-tree dir # find header-level breakage
--magic-fix dir --apply # auto-fix ext/magic mismatches
--bulk-validate dir # run every per-format validator
# then re-run --validate-X on individual failures for detail
This is the 12th cross-format utility — depth-checking that
catches per-format semantic errors (duplicate ids, invalid
enums, contradictory flag combos, dangling cross-refs) that
--audit-tree's header-only scan can't see. CLI flag count
950 -> 951.
