v0.2.0

Changed

• Updated dependency and build support for numpy>=2,<3.
• Dropped support for Python 3.8 and 3.9; supported Python versions are now
  3.10, 3.11, 3.12, and 3.13.
• Reworked genotype backends so the default magenpy backend uses native
  BED-backed C++ kernels, reducing reliance on the pandas-plink / xarray
  stack for core genotype operations.
• Promoted the bed-reader backend to a fully supported backend with coverage
  for genotype extraction, sparse and dense conversion, variant statistics,
  scoring, GWAS, and LD computation.
• Renamed command-line scripts from magenpy_ld and magenpy_simulate to
  mgp_compute_ld and mgp_simulate, and standardized CLI logging through
  the package logger interface.
• Updated GitHub Actions CI and wheel workflows, including newer macOS runner
  versions and Python/Numpy build targets.
• Improved documentation across installation, command-line usage, genotype
  backends, LD workflows, citation, and container-based usage.

Added

• Added native C++/Cython code paths for PLINK BED-backed genotype extraction,
  variant statistics, quantitative-trait GWAS, and score/statistical kernels.
• Added three new command-line utilities for interacting with pre-computed
  magenpy LD matrices:
  • mgp_extract_ld for extracting dense LD submatrices by SNP list or genomic
    region.
  • mgp_prune_ld for LD-based pruning of variant lists or harmonized summary
    statistics.
  • mgp_expand_ld for expanding focal SNP lists to include LD neighbors.
• Added support for reading local ZipStore-backed LD matrices and for streaming
  LD matrices hosted on Hugging Face.
• Added a DockerHub publishing workflow and refreshed Docker CLI container
  support, including PLINK/PLINK2 availability and validation of all CLI tools.
• Added Apptainer documentation for running the DockerHub image on shared
  computing systems where Docker is unavailable.
• Added and expanded tests for the native magenpy backend, bed-reader
  backend, PLINK backend comparisons, LD matrix handling, CLI scripts, and
  summary-statistics parsers.
• Added uv-based manual testing instructions and refreshed package metadata
  for the 0.2 release.
• Added citation and AI declaration documentation.

Fixed

• Improved robustness of GWAS summary-statistics parsers, including support for
  additional column-name conventions and stricter parser tests.
• Added support for writing summary statistics in supported output formats,
  making the parser layer more useful for format conversion workflows.
• Improved robustness of LD matrix creation and conversion from sparse inputs,
  including handling of row-wise gaps in the upper-triangular LD representation.
• Fixed and improved phenotype simulation, xarrayGenotypeMatrix, and utility
  behavior encountered during the Python/Numpy/backend upgrade work.
• Improved command-line tests and documentation so installed scripts, Docker
  checks, and documentation now use the same mgp_* command names.