lic: update license rules for cdla permissive 2.0

[davidgamez]

Description

This PR updates the resource location to support license categorization in the Python published package.

This pull request introduces several improvements and new features to the license classification tooling, focusing on enhanced automation, better documentation, and improved package data handling. The most significant changes include the addition of a detailed script and documentation for LLM-based license classification, updates to the packaging configuration to ensure necessary data files are included, and enhancements to the classification schema for the CDLA-Permissive-2.0 license.

Key changes:

1. License Classifier Automation and Documentation

• Added comprehensive documentation for the new classify_license.sh script, which uses an LLM (default: gpt-4.1) to classify licenses and write standardized permissions, conditions, limitations, tags, and reasons fields to license JSON files. The documentation covers usage, options, few-shot example injection, and environment variables.
• Updated the classify_license.sh script help text to clarify that the default system/user prompt files are now bundled package data, not in docs/. Added documentation for the new --max-examples flag, which controls the number of few-shot examples injected into the LLM prompt. [1] [2]

2. Packaging Improvements

• Modified pyproject.toml to ensure that all necessary data files (JSON, Markdown, and example JSON files) are included in the licensing.classify package, supporting out-of-tree script and package usage.

3. License Classification Schema Enhancement

• Updated the CDLA-Permissive-2.0.json license data to add detailed permissions, conditions, limitations, and tags arrays, as well as a comprehensive reasons block with verbatim and inferred evidence for each rule, improving the accuracy and transparency of license classification.

4. Obsolete Documentation Removal

• Removed the now-obsolete docs/classification/SYSTEM_PROMPT.md file, as the system prompt is now bundled with the package and referenced accordingly in scripts and documentation.