When designing a Bigtable schema with several groups of columns that are frequently used together, how should you organize these groups to optimize read performance according to Google Cloud's best practices?