Self-Supervised Speech Quality Estimation and Enhancement Using Only Clean Speech