Handle external_data table deploys mostly the same as for regular tables

[curtismorales]

Description

Fixes a bug introduced in #6529 -- this PR added support for tables that are defined with a metadata.yaml but no query.sql, but didn't check for tables that are built on external data (Google Sheets or CSV)

This adds logic to the deploy_table function to handle external tables differently. In the process I also did some refactoring of how we deploy external tables. I moved the metadata update logic that was in _deploy_external_data into an attach_external_data_config function in publish_metadata.py and removed the duplicated code.

Is this reasonable? This also cuts one of the passes we're doing through all the sql and metadata files during a deploy, so I think it's a performance benefit. Unless there was a reason that we weren't deploying external tables in the same threads as the rest of the table deploys?

Also, I should add some tests; this bug is kind of my fault for not adding tests in the first place. If it's okay with you, I'd like to do that in a separate PR.

Related Tickets & Documents

• AD-501 (where this bug came up)

┆Issue is synchronized with this Jira Task

[scholtzan]

Should we check for other types of query files, like query.py here as well. Maybe similar to

bigquery-etl/bigquery_etl/cli/query.py

Lines 2222 to 2226 in 5d216fa

	and not any(
	file.suffix == ".sql" or file.name == "query.py"
	for file in metadata_file.parent.iterdir()
	if file.is_file()
	)

[scholtzan]

Having some more unit tests for this would be nice, but it's fine to do that in a follow-up PR.