We need a stronger No_GC_Verifier since the current one only detects that a GC occurred, not whether it's possible for a GC to occur.

I think the logic would have to be something like counting the number
of times we do a safepoint check and seeing if that number changes
while the No_GC_Verifier is on the stack. It would also have to take
into account whether the GC_locker was active or not to determine
whether even that even if a safepoint occurred it would have been
suppressed. It would also have to determine if you transitioned to or
were in native thread state.

###@###.### 2003-05-15