Add ability to insert distributed columns at Pandas query compiler

[dchigarev]

For now, there is no way to insert one or multiple columns in a dataframe at a specific location if you are working at the query compiler level, without materializing column and passing it into insert.

[devin-petersohn]

I agree, this can use something better than what we have. Right now, I would say use concat:

return query_compiler.get_column_array(self.columns[:loc]).concat(1, new_value).concat(1, query_compiler.get_column_array(self.columns[loc:]))

No guarantee on it working as I wrote it 😄.