Pre-release

Added type checking and automatic linting/formatting, #97
Added a utility function for de-structuring version strings in oqs.py
- version(version_str: str) -> tuple[str, str, str]: - Returns a tuple containing the (major, minor, patch) versions
A warning is issued only if the liboqs-python version's major and minor numbers differ from those of liboqs, ignoring the patch version

Version 0.12.0 - January 15, 2025

Fixes #98. The API that NIST has introduced in FIPS 204 for ML-DSA includes a context string of length >= 0. Added new API for signing with a context string
- Signature.sign_with_ctx_str(self, message, context)
- Signature.verify_with_ctx_str(self, message, signature, context, public_key)
When operations fail (i.e., OQS_SUCCESS != 0) in functions returning non-boolean objects, a RuntimeError is now raised, instead of returning 0
Bugfix on Linux platforms, c_int -> c_size_t for buffer sizes
Pyright type checking fixes
Updated examples to use ML-KEM and ML-DSA as the defaults

Replaced CHANGES by CHANGES.md, as we now use Markdown format to keep track of changes in new releases
Removed the NIST PRNG as the latter is no longer exposed by liboqs' public API
liboqs is installed automatically if it is not detected at runtime

This is a maintenance release, minor deprecation fixes
Python minimum required version is enforced to Python 3.8 in pyproject.toml
To follow Python conventions, renamed in oqs/oqs.py:
- is_KEM_enabled() -> is_kem_enabled()
- get_enabled_KEM_mechanisms() -> get_enabled_kem_mechanisms()
- get_supported_KEM_mechanisms() -> get_supported_kem_mechanisms()

This is a maintenance release, minor fixes
Minimalistic Docker support
Migrated installation method to pyproject.toml
Removed AppVeyor and CircleCI, all continuous integration is now done via GitHub actions

Added library version retrieval functions:
- oqs_version()
- oqs_python_version()

Release numbering updated to match liboqs
Added macOS support on CircleCI, we now support macOS & Linux (CircleCI) and Windows (AppVeyor)

The liboqs handle has now module-private visibility in oqs.py so clients can not access it directly; can be accessed via the new oqs.native() function
Closing #7 link, all issues addressed
Added AppVeyor continuous integration

This release updates for compatibility with liboqs 0.2.0, which contains new/updated algorithms based on NIST Round 2 submissions.