Replace pandas.DataFrame with PyArrow.Table for nullable int typing (#8733)

* Use PyArrow Table for query result serialization

* Cleanup dev comments

* Additional cleanup

* WIP: tests

* Remove explicit dtype logic from db_engine_specs

* Remove obsolete  column property

* SupersetTable column types

* Port SupersetDataFrame methods to SupersetTable

* Add test for nullable boolean columns

* Support datetime values with timezone offsets

* Black formatting

* Pylint

* More linting/formatting

* Resolve issue with timezones not appearing in results

* Types

* Enable running of tests in tests/db_engine_specs

* Resolve application context errors

* Refactor and add tests for pyodbc.Row conversion

* Appease isort, regardless of isort:skip

* Re-enable RESULTS_BACKEND_USE_MSGPACK default based on benchmarks

* Dataframe typing and nits

* Renames to reduce ambiguity

superset/assets/spec/javascripts/sqllab/ExploreResultsButton_spec.jsx

@@ -55,13 +55,11 @@ describe('ExploreResultsButton', () => {
   const mockColumns = {
     ds: {
       is_date: true,
-      is_dim: false,
       name: 'ds',
       type: 'STRING',
     },
     gender: {
       is_date: false,
-      is_dim: true,
       name: 'gender',
       type: 'STRING',
     },