1. How does a database's transaction isolation level influence the potential for phenomena such as dirty reads, non-repeatable reads, and phantom reads, and what are the trade-offs between maintaining strict isolation and achieving high performance in a multi-user environment ?