fix group-by v2 BufferArrayGrouper for empty multi-value dimension row

[clintropolis]

This PR fixes an 'off by 1' issue in BufferArrayGrouper for the scenario that when faced with multi-value dimension rows with 0 elements, and attempting to aggregate on GroupByColumnSelectorStrategy.GROUP_BY_MISSING_VALUE, the grouper would initialize incorrectly into a state where it's next value was -1, meaning hasNext evaluates to false, but next is called without checking resulting in something like:

java.lang.RuntimeException: java.util.NoSuchElementException
	at org.apache.druid.query.aggregation.AggregationTestHelper$9.run(AggregationTestHelper.java:678) ~[test-classes/:?]
	at org.apache.druid.query.groupby.epinephelinae.GroupByMergingQueryRunnerV2$1$1$1.call(GroupByMergingQueryRunnerV2.java:243) [classes/:?]
	at org.apache.druid.query.groupby.epinephelinae.GroupByMergingQueryRunnerV2$1$1$1.call(GroupByMergingQueryRunnerV2.java:232) [classes/:?]
	at java.util.concurrent.FutureTask.run(FutureTask.java:266) [?:1.8.0_192]
	at org.apache.druid.java.util.common.concurrent.DirectExecutorService.execute(DirectExecutorService.java:82) [classes/:?]
	at java.util.concurrent.AbstractExecutorService.submit(AbstractExecutorService.java:134) [?:1.8.0_192]
	at com.google.common.util.concurrent.AbstractListeningExecutorService.submit(AbstractListeningExecutorService.java:58) [guava-16.0.1.jar:?]
	at org.apache.druid.query.groupby.epinephelinae.GroupByMergingQueryRunnerV2$1$1.apply(GroupByMergingQueryRunnerV2.java:230) [classes/:?]
	at org.apache.druid.query.groupby.epinephelinae.GroupByMergingQueryRunnerV2$1$1.apply(GroupByMergingQueryRunnerV2.java:220) [classes/:?]
	at com.google.common.collect.Iterators$8.transform(Iterators.java:794) [guava-16.0.1.jar:?]
	at com.google.common.collect.TransformedIterator.next(TransformedIterator.java:48) [guava-16.0.1.jar:?]
	at com.google.common.collect.Iterators.addAll(Iterators.java:357) [guava-16.0.1.jar:?]
	at com.google.common.collect.Lists.newArrayList(Lists.java:147) [guava-16.0.1.jar:?]
	at com.google.common.collect.Lists.newArrayList(Lists.java:129) [guava-16.0.1.jar:?]
	at org.apache.druid.query.groupby.epinephelinae.GroupByMergingQueryRunnerV2$1.make(GroupByMergingQueryRunnerV2.java:216) [classes/:?]
	at org.apache.druid.query.groupby.epinephelinae.GroupByMergingQueryRunnerV2$1.make(GroupByMergingQueryRunnerV2.java:158) [classes/:?]
...

Also fixes an issue in GroupByEngineIterator#hasNext to return delegate.hasNext instead of assuming the delegate has a next after initialized. This is no longer an issue since the delegate in this case is now getting initialized correctly, but still seems like the right thing to do.

Finally, fixes a third issue with GroupByQueryEngineV2.ArrayAggregateIterator and sql compatible null handling, where it was always using "" as the key for GroupByColumnSelectorStrategy.GROUP_BY_MISSING_VALUE instead of what the group by strategies for the hash iterator was doing for string dimensions and using NullHandling.defaultStringValue().

The added test will trigger the NoSuchElementException without the modifications in this PR (test code was borrowed from #7588). Additionally, the tests are run with both the array grouper and hash grouper to ensure they produce the same results (the hash grouper did not have this issue). The tests would also fail for sql compatible null handling for the array grouper without the 3rd fix to ArrayAggregateIterator, which is how I stumbled into that issue.

[jihoonson]

Would you please check the CI failure?

[clintropolis]

Would you please check the CI failure?

Yeah, unfortunately the CI failure is legitimate, I'm trying to determine the issue. Interestingly, this appears to be another inconsistency between BufferHashGrouper and BufferArrayGrouper, as only the latter is failing this test.. though I haven't really determined how that is possible yet since neither of them directly do anything with NullHandling stuff.

[jihoonson]

LGTM. Thanks @clintropolis.