v0.2.1 - Performance improvement and code quality

What's Changed

Performance

• fit() / fit_period() and transform() now use pandas vectorized groupby
  instead of row-by-row Python loops (itertuples), significantly reducing
  processing time on large interaction logs.

Documentation

• Docstrings for RecencyFrequencyScorer and RFOptimizer have been improved
  for clarity and IDE support.
• docs/functional-design.md and docs/development-guidelines.md updated to
  reflect fit_period() alongside fit() in all relevant descriptions.

Tests

• Added comprehensive TestEvaluate class covering precision, recall, f1,
  recall_norm, and f1_norm metrics (129 tests total).
• Fixed test_resets_solve_state to correctly verify solver state reset after
  build_model().

Internal

• String quotes unified to double quotes throughout scorer.py and
  optimizer.py.

Full Changelog

v0.2.0...v0.2.1