Add support for Thread.onSpinWait() while spinning

[franz1981]

It wides the scope of #147, using it on queues by reusing the ideas on https://github.com/giltene/GilExamples/tree/master/SpinWaitTest/src/main/java/org/performancehints to provide support while retaining Java 6 compatibility.

[franz1981]

Just for the records, I've run a couple of benchmarks using onSpinWait using this branch :

6P-6C (4 chunks of 256 elements):

Benchmark                                         (burstSize)  (chunkSize)  (consumerCount)  (onSpinWait)  (pooledChunks)  (qCapacity)                      (qType)  (warmup)  Mode  Cnt       Score       Error   Units
XaddQueueBurstCost.burstCost                              100          256                6          true               4         1024  MpmcUnboundedXaddArrayQueue      true  avgt   30  183017.519 ± 10334.573   ns/op
XaddQueueBurstCost.burstCost                              100          256                6         false               4         1024  MpmcUnboundedXaddArrayQueue      true  avgt   30  191742.466 ± 13786.168   ns/op

10P-1C (6 chunks of 256 elements):

Benchmark                                         (burstSize)  (chunkSize)  (consumerCount)  (onSpinWait)  (pooledChunks)  (qCapacity)                      (qType)  (warmup)  Mode  Cnt      Score      Error   Units
XaddQueueBurstCost.burstCost                              100          256                1          true               5       132000  MpscUnboundedXaddArrayQueue      true  avgt   30  51946.455 ±  978.654   ns/op
XaddQueueBurstCost.burstCost                              100          256                1          true               5       132000  MpmcUnboundedXaddArrayQueue      true  avgt   30  65129.730 ± 1625.001   ns/op
XaddQueueBurstCost.burstCost                              100          256                1         false               5       132000  MpscUnboundedXaddArrayQueue      true  avgt   30  53812.430 ± 1962.361   ns/op
XaddQueueBurstCost.burstCost                              100          256                1         false               5       132000  MpmcUnboundedXaddArrayQueue      true  avgt   30  69046.787 ± 3327.661   ns/op

with 1024 elements (as a reference, no onSpinWait here):

Benchmark                                         (burstSize)  (chunkSize)  (consumerCount)  (onSpinWait)  (pooledChunks)  (qCapacity)                  (qType)  (warmup)  Mode  Cnt       Score      Error   Units
XaddQueueBurstCost.burstCost                              100          256                1         false               4         1024  MpscUnboundedArrayQueue      true  avgt   30  233283.374 ± 2846.983   ns/op