(synopsis is provisional, would probably be changed)

This is a follow-up from JDK-8338379, which made sure we are properly synchronizing the class state via class-init-checks. Notable places that check class-init-state are calls to static methods (with the help of clinit-barriers) and allocation (`new`), checking class-init-checks directly. The remaining question is whether C1 and C2 emit class-init-checks properly and/or do not elide them improperly.

I wonder if clinit barriers should actually be done for all methods, regardless of their static-ness. There may be more interesting cases like these.