And also use mm_base_modem_at_command() instead of the full() version,
as we're anyway mm_base_modem_peek_best_at_port().
This commit also fixes the logic where we make sure the finish()
method completes the GTask that is created in the async method
originally.
d1436ac0c6