RANDOM-INVENTORY COMPARISON — integrated in v1.51

Read main Methods 2.10, Results 3.8 and Supplementary Information S11.2.1.
The original standalone audit is retained byte for byte. LEIA_PRIMEIRO.txt
and SOURCE.json describe its original stage and v1.50 source; the statement
that v1.50 was untouched was true at that stage. v1.51 now integrates the
results and preserves this execution history. This is not preregistration.

With fixed Cygnus, roles, quarter-degree calibration and search window,
33/1000 general-uniform and 105/1000 all-even six-count inventories meet
the joint archaeological-interval and observed-RMS criterion. The original
date-informed selection replay gives 24/1000 and 105/1000 respectively.
Any-epoch RMS <0.2 degree occurs in 917/1000 and 882/1000 respectively.
These are frequencies under specified artificial count distributions.
They do not estimate historical intention or correct the entire search history.

Run in a disposable copy, because scripts overwrite result files:
  python -m pip install -r requirements.txt
  python audit_random_inventories.py
  python verify_audit.py

The local SHA256SUMS.txt verifies the 14 original audit files; the parent
archive manifest additionally covers this current guide. Astronomical
model files match those in S11_Woodhenge. SUMMARY.json records both
distributions, Wilson Monte Carlo intervals and two-grid checks.
