Add a synthetic tabular-data generator preserving column distributions

[maziyarpanahi]

Description

Adds a local, dependency-free synthetic tabular-data generator that preserves empirical column distributions and numeric rank correlations while preventing exact source-row copies.

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation update
• Test addition/improvement

Changes Made

• Fit reusable numeric and categorical marginal profiles with pairwise numeric rank correlations.
• Generate deterministic seeded samples using stratified empirical quantiles and a Gaussian rank-correlation model.
• Reject exact source-row fingerprints and fail safely when the available support cannot produce unseen rows.
• Report KS/total-variation marginal fidelity, correlation deltas, an overall score, and copied-row count.

Testing

• Added unit tests covering marginals, correlation fidelity, reproducibility, source-row exclusion, missing values, categorical edge cases, and invalid inputs.
• .venv/bin/python -m pytest tests/ -q — 5,211 passed, 46 skipped.
• make format, make lint, and make format-check.
• Scoped pre-commit hooks.

Documentation

• Added public API docstrings and documented the default fidelity tolerances.
• Updated CHANGELOG.md (not required for this scoped feature PR).

Dependencies

• No new dependencies.

Related Issues

Closes #945