Add scheduled full test suite runs every 6 hours

[kenyonj]

Summary

• Add schedule trigger to run the full test suite every 6 hours
• Scheduled and workflow_dispatch runs execute all test types regardless of file changes
• PR and merge_group events continue using incremental diff-based testing

Changes

• .github/workflows/test.yml:
  • Add schedule trigger with cron: "0 */6 * * *"
  • Add run_type step to detect scheduled/manual runs
  • Skip git diff steps during scheduled runs (no PR context)
  • Run all test types unconditionally during scheduled runs

Why

This ensures collection rename issues are caught regularly (every 6 hours) without blocking every PR with expensive API checks. Combined with the hourly collections-renames workflow, this provides comprehensive coverage while keeping PR CI fast.

[Copilot]

Pull request overview

This PR adds scheduled full test suite runs every 6 hours to complement the existing hourly collection-renames workflow, ensuring comprehensive testing coverage while keeping PR-based CI fast. The implementation uses GitHub Actions' schedule trigger with a cron expression and introduces logic to differentiate between incremental (PR/merge) and full (scheduled/manual) test runs.

Changes:

• Add schedule trigger with cron: "0 */6 * * *" to run tests every 6 hours
• Introduce run type detection that treats schedule and workflow_dispatch events as full runs
• Modify git diff steps to skip during full runs (since there's no PR context)
• Update Ruby setup and test execution conditionals to always run during full runs

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

The current implementation relies on a subtle Ruby behavior that may not be immediately obvious. When git diff steps are skipped during scheduled runs, the environment variables (TOPIC_FILES, COLLECTION_FILES, TEST_ALL_FILES) are set to empty strings. The test helpers check if ENV.fetch("TEST_ALL_FILES", false), and since empty strings are truthy in Ruby, this causes the wildcard pattern to be returned correctly.

However, this behavior is fragile and unclear:

1. It relies on empty strings being truthy in Ruby (only nil and false are falsy)
2. It depends on GitHub Actions evaluating undefined step outputs as empty strings
3. The intended behavior is not explicit in the workflow

Consider making this more explicit and robust by either:

1. Setting TEST_ALL_FILES to a clear sentinel value like "true" or "1" when full_run is true
2. Adding a comment explaining this intentional behavior
3. Using a dedicated full_run environment variable that the test helpers check explicitly