G1: Compare max_parallel_refinement_threads to UINT_MAX

in g1Arguments.cpp:

  // Verify that the maximum parallelism isn't too high to eventually overflow
  // the refcount in G1CardSetContainer.
  uint max_parallel_refinement_threads = G1ConcRefinementThreads + G1DirtyCardQueueSet::num_par_ids();
  uint const divisor = 3; // Safe divisor; we increment by 2 for each claim, but there is a small initial value.
  if (max_parallel_refinement_threads > UINTPTR_MAX / divisor) {
    vm_exit_during_initialization("Too large parallelism for remembered sets.");
  }


max_parallel_refinement_threads will always be less than UINTPTR_MAX on platforms where pointers are larger than unsigned. Use the intended UINT_MAX instead.