[REDO] store/load order not preserved when handling memory pool due to weakly ordered memory architecture of aarch64

Has originally been spotted with:

vmTestbase/vm/mlvm/meth/stress/compiler/deoptimize/Test.java

Seems like it's not the only one:

vmTestbase/vm/mlvm/meth/stress/compiler/sequences/Test.java

Has been seen only in macosx-aarch64 so far. I tried linux, macosx with x64 and aarch64, in both release and debug configurations, only macosx-aarch64 showed the problem (approx. one out of 10 runs). The memory pool in question is 'Non-segmented code cache' (the test was started with '-XX:ReservedCodeCacheSize=100m -XX:+CreateCoredumpOnCrash -XX:TieredStopAtLevel=2').

The stacktrace:

# ERROR: java.lang.InternalError: Memory Pool not found
# ERROR: at java.management/sun.management.MemoryPoolImpl.getUsage0(Native Method)
# ERROR: at java.management/sun.management.MemoryPoolImpl.getUsage(MemoryPoolImpl.java:94)
# ERROR: at vm.mlvm.meth.share.MHTransformationGen$CodeCacheMonitor.lambda$isCodeCacheEffectivelyFull$4(MHTransformationGen.java:118)
# ERROR: at vm.mlvm.meth.share.MHTransformationGen$CodeCacheMonitor.lambda$isCodeCacheEffectivelyFull$5(MHTransformationGen.java:122)
# ERROR: at java.base/java.util.Optional.ifPresent(Optional.java:178)
# ERROR: at vm.mlvm.meth.share.MHTransformationGen$CodeCacheMonitor.isCodeCacheEffectivelyFull(MHTransformationGen.java:122)
# ERROR: at vm.mlvm.meth.share.MHTransformationGen.createSequence(MHTransformationGen.java:157)
# ERROR: at vm.mlvm.meth.share.MHTransformationGen.createAndCallSequence(MHTransformationGen.java:516)
# ERROR: at vm.mlvm.meth.stress.compiler.deoptimize.Test.runThread(Test.java:164)
# ERROR: at vm.mlvm.share.MultiThreadedTest.lambda$run$1(MultiThreadedTest.java:71)
# ERROR: at java.base/java.lang.Thread.run(Thread.java:833)