BUG: tries to insert NaN for source_table_name in GroupedQuery count

[tswast]

When count doesn't exist in a group, some columns end up as NaN. May need to filter these out when inserting rows from dataframe.

TODO:

• Fix possible bug in google-cloud-bigquery. fix: omit NaN values when uploading from insert_rows_from_dataframe googleapis/python-bigquery#170
• Add integration test covering this case. fix: update BigQuery dependencies to fix group-by results handler #64

[tswast]

Filed issue with google-cloud-bigquery: googleapis/python-bigquery#169

This is a general problem with uploading dataframes that use NaN as a NULL indicator. In the meantime we could work around by filtering out NaN values before uploading / converting NaN to None, but I think that's best done in the google-cloud-bigquery library.

[tswast]

Fix pending at googleapis/python-bigquery#170

[tswast]

Fix merged. Waiting on release googleapis/python-bigquery#173

[tswast]

Possible workaround while waiting for BigQuery release: https://stackoverflow.com/a/54403705/101923

df = df.replace({np.nan: None})