What's the best way to deal with large schemas? I have 20+ columns in some of my TableAPI tables and I feel like if I continue to update the schema i'll spend all my time altering columns in 5 places in my stream.