Add the param `stream_results`

[hantmac]

if stream_results = false, pool the rows data into cursor, else call next every time.
Also fix #22

[rad-pat]

Does the current driver implementation of RowIterator represent a server-side cursor? If so, possible the dialect options supports_server_side_cursors and server_side_cursors are of use also?

[hantmac]

Does the current driver implementation of RowIterator represent a server-side cursor? If so, possible the dialect options supports_server_side_cursors and server_side_cursors are of use also?

@rad-pat Because the databend-sqlalchemy use the databend-rust-driver-binding as python driver. So the RowIterator in https://github.com/datafuselabs/bendsql/blob/main/sql/src/rows.rs#L120 has implementation of RowIterator represent a server-side cursor.

[rad-pat]

Actually, it looks like flag server_side_cursors is the old way anyway, now deprecated in favour of stream_results.

[hantmac]

Actually, it looks like flag server_side_cursors is the old way anyway, now deprecated in favour of stream_results.

We prefer to have a stream_results param, and default value is true.

[TCeason]

Now we can already support server_side_cursors https://github.com/datafuselabs/databend-sqlalchemy/blob/main/databend_sqlalchemy/databend_dialect.py#L609 use this demo to replace stream_result.

import connector
cursor = connector.connect('databend://root:@localhost:8000?sslmode=disable').cursor()
cursor.execute('SELECT * FROM test')
# print(cursor.fetchone())
# print(cursor.fetchall())
for row in cursor:
    print(row)