syntheticdata 0.1.0

• Initial release.
• synthesize(): Parametric (Gaussian copula), bootstrap with noise injection, and Laplace noise perturbation methods.
• validate_synthetic(): Distributional similarity (KS), correlation preservation, discriminative accuracy (AUC), and nearest-neighbor privacy ratio.
• compare_methods(): Benchmark all three synthesis methods on the same dataset.
• privacy_risk(): Re-identification risk assessment via nearest-neighbor distance ratio, membership inference, and attribute disclosure.
• model_fidelity(): Train-on-synthetic, test-on-real predictive performance comparison.