Autodiscover dataflow

[jm-rivera]

This pull request introduces ODA Reader version 1.5.0, focusing on a major improvement to dataflow version handling. The old blind version-decrement fallback is replaced by authoritative version discovery using the OECD SDMX metadata endpoint. This ensures the package always finds the latest schema version, improving reliability when OECD updates their datasets. The release also adds a new public API function, clear_version_cache(), to force fresh version lookups, and refactors HTTP primitives to eliminate circular imports. Documentation is updated throughout to reflect these changes.

Version Discovery Improvements

• Replaced the old version-decrement fallback logic with a new system that queries the authoritative SDMX metadata endpoint to discover the latest valid dataflow version for all datasets. This makes version handling more robust and transparent. [1] [2] [3] [4]
• Updated documentation and user-facing text to describe "automatic version discovery" instead of "version fallback," clarifying behavior and user expectations. [1] [2] [3]

API and Public Interface

• Added clear_version_cache() to the public API, allowing users to force a fresh version discovery mid-session if the OECD publishes new versions. Updated documentation and __init__.py exports accordingly. [1] [2] [3] [4]

HTTP and Internal Refactoring

• Moved HTTP primitives (rate limiter, cache session, and HTTP GET functions) into a new module _http_primitives.py to break circular imports and improve maintainability. Updated internal references and re-exported for backward compatibility. [1] [2] [3] [4]

Documentation and Examples

• Updated advanced usage and caching documentation to explain the new version discovery mechanism and how to clear the version cache, including code examples. [1] [2]

Version Bump

• Bumped the package version from 1.4.3 to 1.5.0 in pyproject.toml and CHANGELOG.md to reflect these breaking and feature changes. [1] [2]