How to optimize the JOIN of two large tables (10 and 15 million rows) with subsequent filtering if indexes on JOIN keys already exist?