Add dropColumns as TableOperations / TableSpec

[devinrsmith]

This is in support of #3473, where for adapting implementation purposes, it's cleaner and easier to use dropColumns.

[devinrsmith]

For a bit more context, the SQL query:

my_time = time_table("00:00:01").update_view(
    ["I=ii%10==0?null:ii", "B=ii%11==0?null:ii%3==0"]
)

SELECT 
  * 
FROM 
  my_time 
ORDER BY "B", "Timestamp" DESC

produces the logical plan

LogicalSort(sort0=[$2], sort1=[$0], dir0=[ASC], dir1=[DESC])
  LogicalProject(Timestamp=[$0], I=[$1], B=[$2])
    LogicalTableScan(table=[[my_time]])

notice that the hierarchy produced has sort being executed after the project, yet still refers to the columns by reference.

With the DHC implementation, I'd like to adapt all of the tables into "reference" space, execute the plan, and then remove the extra references.

In this implementation, instead of using a view for the projection, I'm able to use an update_view so that the sort can still proceed in reference space, and then finally by dropping the references as the last step.

digraph {
"op_0" ["label"="ticketTable(scan/my_time)"]
"op_1" ["label"="view(ref_0=Timestamp,ref_1=I,ref_2=B)"]
"op_2" ["label"="updateView(Timestamp=ref_0,I=ref_1,B=ref_2)"]
"op_3" ["label"="sort([ASCENDING(ref_2),DESCENDING(ref_0)])"]
"op_4" ["label"="dropColumns([ref_0,ref_1,ref_2])"]
"op_1" -> "op_0"
"op_2" -> "op_1"
"op_3" -> "op_2"
"op_4" -> "op_3"
}

[devinrsmith]

I'd potentially like to follow this up with rename_columns.

[devinrsmith]

So - my current SQL implementation does not rely on dropColumns anymore. That said, it might be nice to get these improvements over the line regardless?

[rcaudy]

Caught up to head, still need to discuss open issue.