Optimization for Refine on staging
- Parameterize execution timeout and retries for faster feedback
- Customize Skein resources for local Spark to avoid resource errors
- Create meta_id root column for better diffing output
- Clean code around skipping _schema column at diffing