What is Snowflake's recommended approach for defining a clustering key on a table with a high-cardinality column, such as an ID_NUMBER column containing 15-digit numeric identifiers?