ts_control, ts_transport_derp: periodic DERP latency checking

Currently, we only check latencies to DERP regions once on startup, and cut off the latency checking after we reach a threshold of responses or a certain amount of time has expired. This cut-off is necessary to keep startup time reasonable - depending on distance and network weather, some DERP regions can take seconds to respond.

Once we're up and connected, tailscale-rs needs to periodically re-check all DERP region latencies, change DERP home regions if necessary, and report new latencies/new home region to the control plane. As we're not constrained by startup time after the first latency check, there shouldn't be a problem reporting all latencies, as we're not in a rush.

Start by locating the ts_control and ts_transport_derp components named in the issue and read how the initial DERP latency check and control-plane reporting work. Trace the existing home-region selection and latency lifecycle before deciding where periodic checks belong. Done means all DERP regions are rechecked after startup, the home region can change, and updated latencies and home-region information reach the control plane.