Support array_union scalar expr

[dharanad]

Which issue does this PR close?

Part of #1042

Rationale for this change

What changes are included in this PR?

How are these changes tested?

[dharanad]

There is this one case which i haven't added.

          checkSparkAnswerAndOperator(
            sql("SELECT array_union(array(_2 * 1), array(_9, _10)) from t1"))

Because it was failing due to mismatch in ordering between spark and df
Example

  == Results ==
  !== Correct Answer - 1000 ==                                        == Spark Answer - 1000 ==
   struct<array_union(array((_2 * -1)), array(_9, _10)):array<int>>   struct<array_union(array((_2 * -1)), array(_9, _10)):array<int>>
  ![ArraySeq(128, -128, -896)]                                        [ArraySeq(128, -896, -128)]

I am not sure how to deal with this, open to suggestions

[andygrove]

I'm closing this issue because it has been stale for a while. @dharanad feel free to reopen this if you resume work on this featrure.