More of a graphql question but for the query situations that don't work well from Prisma side (aggregations like SUM, AVG, etc), is there a generally accepted approach of either writing raw sql in your resolver or using an ORM?