Inefficient algorithm for converting JMX attribute names to native resource attribute names

ModelControllerMBeanHelper is taking an MRR, using it to create a DMR ModelNode description of the resource type, and then iterating over the attribute names in that description tree to try and match up with the requested JMX attribute name. That middle step of creating the DMR description is very expensive; it produces a very large node tree. It's not necessary; the MRR already can tell you the legal attribute names.

I believe this is legacy cruft from the early AS 7 days when the MRR wasn't always reliable for all attribute names. But it's been reliable for a long time now.