On Tanaka‐Mura's fatigue crack nucleation model and validation