Seeded calls restore the caller's random number state on exit, and seed
defaults to NULL in heldout_test_ordpp(), permutation_ordpp() and
stability_ordpp().
Default maxit raised to 500: item-specific fits often need more than 150.
simulate_ordpp() gains strength for weak-dependence power studies.
Characteristic functions are evaluated on distinct values (much faster).
New index = "shape" option compares the variance-standardized projection
with the independence reference, removing the pairwise-covariance component.
New decompose_ordpp() reports Var(Z)/Var(Z0) with the full and shape indices.
New heldout_test_ordpp(): exact permutation test along a frozen map.
permutation_ordpp() accepts fit = so the null fits reuse its exact settings.
New stability_ordpp() for (cluster) bootstrap stability of loadings and scores.
New as_ordpp_map() wraps external scores/loadings (e.g. optimal-scaling PCA)
for evaluation with the same held-out tools.
simulate_ordpp() gains scenario = ("independent", "rare", "two_factor",
"response_style").
Fitted objects store transformed data, weights, settings, the call and the number of gap parameters on their bounds.