Store grouped-only regression baselines and run slow cases in update workflow

[harryswift01]

Summary

This PR updates the regression baseline workflow so that regenerated baselines only store the grouped result data used by the regression comparison.

It also updates the GitHub workflow for baseline regeneration to include --run-slow, ensuring slow regression cases are included when updating stored baselines.

Changes

Regression baseline output

• Added a baseline payload helper to persist only the groups section of the CodeEntropy output.
• Removed runtime arguments and provenance metadata from regenerated baseline files.
• Keeps normal CodeEntropy output unchanged for users.
• Keeps regression comparison focused on the grouped entropy values that are actually tested.

GitHub baseline update workflow

• Added --run-slow to the baseline update command.
• Ensures slow-marked regression tests are included when regenerating baselines in CI.
• Avoids accidentally updating only the faster subset of regression baselines.

Impact

• Regression baseline files are cleaner and more stable.
• Machine-specific data such as temporary paths, Python versions, platform details, and Git SHAs will no longer be committed in updated baselines.
• CI baseline updates now cover the full intended regression suite, including slow cases.
• User-facing JSON output is unaffected.

[jimboid]

Looks good, improves the testing base lines workflow.