Test to determine if klass is task reentrant is incorrect in implementation of
bytecode new on x86.
This results in execution never taking the "task-reentrant" case branch and
systematically in the interpreter runtime, which impacts the performance of the interpreter.
Luckily, this doesn't do more harm than poor performance of the "new" bytecode
as the class initialization test in the non-reentrant case when performed
agaisnt a task-reentrant class always fails, routing execution to slow case
systematically.