* persist statistics somewhere inside the execution engine (i.e., persist base table statistics in datafusion catalog?) * implement analyze command + executor to compute base table statistics