Optimize label translation to skip already-translated labels

[KallynGowdy]

Summary

This change optimizes the documentation translation process by only translating newly added labels, rather than re-translating the entire labels file on every run. Previously, all labels were sent for translation each time, even if they hadn't changed.

Key Changes

• Modified translateLabels() function to accept an existing parameter containing previously translated labels, allowing the function to skip re-translating unchanged labels
• Updated label collection logic to only gather strings that don't already have translations, reducing API calls and translation costs
• Added countMissingLabels() helper function to count how many labels need translation, enabling better user feedback
• Improved staleness detection in the main translation flow to determine when labels need re-translation:
  • If the English labels file is newer than the translation output, all labels are re-translated (in case existing labels were reworded)
  • If the translation is up-to-date, existing translations are reused and only new labels are translated
  • The --force flag still triggers a full re-translation
• Enhanced console output to show how many labels were translated, providing visibility into the translation work performed

Implementation Details

• The collect() and rebuild() functions now traverse both the current labels and existing translations in parallel, comparing keys at each level
• Empty translation requests are handled gracefully (no API call if no new labels need translation)
• The logic preserves the hierarchical structure of labels while selectively translating only new entries

https://claude.ai/code/session_01YMDJWsX8SQrGwpxLydExuZ