Active Admittance Control with Iterative Learning for General-Purpose Contact-Rich Manipulation