[Graal] nsk/jvmti/unit/ForceEarlyReturn/earlyretbase should not suspend the thread when the top frame executes JVMCI code

The test creates an instance of inner class earlyretThread and starts it as a separate java thread. Then it forces early return from the thread's method activeMethod(). Before doing a force return the test suspends the target thread. However, the test doesn't check that after the target thread is suspended activeMethod() is executed in the top frame. It might not be the case when the invocation counter overflows and triggers a compilation that results in the top frame started executing some JVMCI code. If this happens the test makes force early return for a wrong method and fails.