zidstats: two-sample testing, ranking, and distributional diagnostics
zidstats 0.2.2 This performance release preserves the public API and statistical definitions of zidstats 0.2.1. Vectorizes the common no-tie RISE graph construction while retaining the label-invariant tie-averaged extension. Computes permutation within-sample sums in batches and reuses internally validated pair-weight matrices. Uses SciPy's compiled sparse MST implementation for ordinary GET when edge weights are strictly ordered. The optional get-accel extra further compiles the complete successive-MST sequence as one ordered-edge cascade. Tied inputs retain the explicit uniqueness checks and existing error behavior. Evaluates named symmetric dissimilarities through a condensed pairwise calculation rather than a full two-direction distance call. Regression tests verify exact or floating-point-equivalent results against the 0.2.1 implementation, including tied RISE inputs and the complete ordinary-GET edge union. The release artifacts pass the full test suite, distribution smoke tests, archive/source identity checks, and twine check under Python 3.12.