Make it easier to debug sqlparam issues with seeds

[jtcohen6]

dbt uses SQL parameters when inserting static data into seeds. The most common SQL parameter is %s, but pyodbc uses ? instead. Today, we achieve that switch at the very last moment, right when dbt is going to execute the SQL:

dbt-spark/dbt/adapters/spark/connections.py

Lines 268 to 271 in 1f84005

	# pyodbc only supports `qmark` sql params!
	query = sqlparams.SQLParams('format', 'qmark')
	sql, bindings = query.format(sql, bindings)
	self._cursor.execute(sql, *bindings)

This is tricky, though, because it means that the SQL logged to logs/dbt.log appears to contain the incorrect SQL parameters. I'd much rather have it be the correct one, if possible, to aid in debugging what dbt actually ran.

We refactored the seed materialization for v0.21, and kicked get_binding_char() into its own macro. I think this means that we could:

• reimplement spark__get_binding_char to return %s or ? based on target.method == 'odbc'
• reimplement spark__load_csv_rows to look more like default__load_csv_rows
• do we still need a custom seed materialization at all?