Add udf compiler unit tests (NVIDIA#583)

* Add udf compiler unit tests

- This change includes unit tests for implemented udf opcodes

Signed-off-by: Allen Xu <allxu@nvidia.com>

Co-authored-by: Sean Lee <selee@nvidia.com>
Co-authored-by: Nicholas Edelman <nedelman@nvidia.com>
Co-authored-by: Alessandro Bellina <abellina@nvidia.com>

* Remove extra space lines in tests

Signed-off-by: Allen Xu <allxu@nvidia.com>

* Remove irrelevant comment

Signed-off-by: Sean Lee <selee@nvidia.com>

* Get the store index correctly

Signed-off-by: Sean Lee <selee@nvidia.com>

* Fix the ldc test

Signed-off-by: Sean Lee <selee@nvidia.com>

* Mark un-compiled test cases

Signed-off-by: Allen Xu <allxu@nvidia.com>

* Rename fall-back test cases

Signed-off-by: Allen Xu <allxu@nvidia.com>

* Fix column comparison

x.except(y) doesn't actually perform element-wise comparison
between x and y.  It returns a new dataframe with the rows from
x that do not appear in y.
It was producing false-positives with some unit tests.

Signed-off-by: Sean Lee <selee@nvidia.com>

* Improve the column comparison performance further

Signed-off-by: Sean Lee <selee@nvidia.com>

Co-authored-by: Allen Xu <allxu@nvidia.com>
Co-authored-by: Sean Lee <selee@nvidia.com>
Co-authored-by: Nicholas Edelman <nedelman@nvidia.com>
Co-authored-by: Alessandro Bellina <abellina@nvidia.com>

udf-compiler/src/main/scala/com/nvidia/spark/udf/Instruction.scala

@@ -528,7 +528,9 @@ object Instruction {
     val opcode: Int = codeIterator.byteAt(offset)
     val operand: Int = opcode match {
       case Opcode.ALOAD | Opcode.DLOAD | Opcode.FLOAD |
-           Opcode.ILOAD | Opcode.LLOAD | Opcode.LDC =>
+           Opcode.ILOAD | Opcode.LLOAD | Opcode.LDC |
+           Opcode.ASTORE | Opcode.DSTORE | Opcode.FSTORE |
+           Opcode.ISTORE | Opcode.LSTORE =>
         codeIterator.byteAt(offset + 1)
       case Opcode.BIPUSH =>
         codeIterator.signedByteAt(offset + 1)