[Java] Avro - Experiment with "compiled" consumer delegates for performance.

All consumers that rely on delegates (e.g. struct, composite, list, union, ...) require  megamorphic lookups which can't be inlined well by JIT.  

We should verify the performance different of a hand-coded consumer vs an existing delegate consumer

i.e. something like:

void consume(Decoder d) {

  ((IntConsumer)delegate).consume(d);

}

compared to the existing implementation.  It is expected we will see a decent amount of performance improvement from this approach.  If we do, we should add an option to converter to generate new custom classes on the fly, that mimic the hand-coded option.

**Reporter**: [Micah Kornfield](https://issues.apache.org/jira/browse/ARROW-6595) / @emkornfield

_{**Note**: *This issue was originally created as [ARROW-6595](https://issues.apache.org/jira/browse/ARROW-6595). Please see the [migration documentation](https://github.com/apache/arrow/issues/14542) for further details.*}

Start by locating the delegate-based consumers and the hand-coded consumer path described in the issue. Compare their performance, then assess whether converter-generated custom classes are warranted; done means a documented comparison and a scoped implementation decision.

Written by the indexing model from the issue text.