Add GA-RFD

[py-cyber]

This pull request introduces a genetic algorithm for the discovery of relaxed functional dependencies (GA‑RFD)

New GA‑RFD Discovery Algorithm

• src/core/algorithms/rfd/ga_rfd/ga_rfd.h and src/core/algorithms/rfd/ga_rfd/ga_rfd.cpp
  Core implementation of the GaRfd class. It builds difference‑dataset bit matrices, maintains an LRU cache for support computation and evolves a population of candidate RFDs using selection, crossover, and mutation operators. The algorithm works on a single table, stores data column‑wise for performance, and outputs a set of unique RFDs with their conf and supp

• src/core/algorithms/rfd/rfd.h
  Lightweight RFD struct (LHS mask, RHS index, confidence, support) with ordering for deduplication;

• src/core/algorithms/rfd/similarity_metric.h and src/core/algorithms/rfd/similarity_metric.cpp
  Abstract SimilarityMetric interface and built‑in metrics. Users can provide custom metrics from Python.

Python Bindings

• src/python_bindings/rfd/bind_ga_rfd.cpp and src/python_bindings/rfd/bind_ga_rfd.h
  Expose GaRfd, RFD, and SimilarityMetric to Python. A helper PySimilarityMetric wraps Python callables so that user‑defined metrics can be passed directly;

• src/python_bindings/rfd/py_similarity_metric.h
  Python‑friendly metric adapter with GIL handling;

• src/python_bindings/py_util/py_to_any.cpp
  Registers a converter for std::shared_ptr<SimilarityMetric> to enable passing Python functions as metrics;

• Example
  examples/basic/mining_ga_rfd.py demonstrates loading Iris, configuring metrics, setting GA parameters, and printing the discovered RFDs.

Tests

• src/tests/unit/test_ga_rfd.cpp
  Extensive test suite covering:
  • Exceptions on empty tables, single attributes, too many attributes, and single rows;
  • Parameterised tests on various datasets (Iris, BreastCancer, Neighbors10k, TestFD, TestLong, TestWide, etc.);
  • Correctness checks (LHS ≠ 0, RHS not in LHS, confidence ≥ ε);
  • Performance benchmarks (Neighbors10k with both fast and slow settings);
  • Determinism with fixed seed;
  • Levenshtein and equality metric variants.

Help links:
report
FlameGraph
paper