[BUG] TPC-DS query 6 now produces incorrect results.

[revans2]

Describe the bug
When I run TPC-DS query 6 on scale factor 1 I get an incorrect result relative to the CPU. The GPU version is missing a null state count.

Steps/Code to reproduce bug

./run_pyspark_from_build.sh -k 'tpcds[q6]' --tpcds_path=/home/roberte/src/data/tpcds/SF1_parquet/ --tpcds_format=parquet

Expected behavior
It should pass, but it fails.

Environment details (please complete the following information)
0.2 pre-release

Additional context
I had to make some minor changes to get this to work because the java/scala APIs changed a bit and things needed to be updated. I'll add in a patch.

[revans2]

diff.patch.txt

[revans2]

Looks like df00904 introduced the issue according to git bisect

[revans2]

I just verified that reverting the above change fixes the regressions. I will try to come up with a patch that fixes the issue instead of removing it, but if I cannot get a patch ready in under an hour I'll put up the revert patch.