Cooperative Learning with Gaussian Processes for Euler-Lagrange Systems Tracking Control under Switching Topologies