Better logging/reset mechanism for MarkPool

It is possible for MarkPool instances to leak, which often be difficult to track down and can lead to denial of service with API users. We need a mechanism to ensure MarkPools instances are never leaked, and to also discover bugs that would cause them to leak.

See the folloiwng thread for a more detailed discussion of the problems: https://lists.apache.org/thread.html/914d5e7e0e4bf9001cb048aaaa034608b1b6bcffa04de9a10158302e@%3Cdev.daffodil.apache.org%3E