`read_gbq(table_id, columns=[list of columns])` should actually filter the amount of columns downloaded from the API

[tswast]

Is your feature request related to a problem? Please describe.

Currently, one only uses the columns parameter to re-order the list of columns and it has to exactly match the columns provided in the query or table. See this TODO:

python-bigquery-pandas/pandas_gbq/gbq.py

Lines 939 to 944 in 912b615

	# TODO(kiraksi): allow columns to be a subset of all columns in the table, with follow up PR
	if columns is not None:
	if sorted(columns) == sorted(final_df.columns):
	final_df = final_df[columns]
	else:
	raise InvalidColumnOrder("Column order does not match this DataFrame.")

Describe the solution you'd like

Only download the selected columns if the user passes a list of columns to read_gbq

For queries:

Maybe these still need to have the columns match since one can specify these in SQL? I don't see a selected_fields option in https://cloud.google.com/python/docs/reference/bigquery/latest/google.cloud.bigquery.client.Client#google_cloud_bigquery_client_Client_query_and_wait

For table IDs:

Pass the list of columns through as selected_fields to https://cloud.google.com/python/docs/reference/bigquery/latest/google.cloud.bigquery.client.Client#google_cloud_bigquery_client_Client_list_rows

Starting here:

python-bigquery-pandas/pandas_gbq/gbq.py

Lines 914 to 919 in 912b615

	final_df = connector.download_table(
	query_or_table,
	max_results=max_results,
	progress_bar_type=progress_bar_type,
	dtypes=dtypes,
	)

going through to

python-bigquery-pandas/pandas_gbq/gbq.py

Line 396 in 912b615

rows_iter = self.client.list_rows(table_ref, max_results=max_results)

Describe alternatives you've considered

A clear and concise description of any alternative solutions or features you've considered.

Additional context

Aside: https://googleapis.dev/python/pandas-gbq/latest/reading.html has no mention that a table ID is supported. We should add a sample there.