Improve Unsol Loss detection

[christophe-calmejane]

The algorithm may cause some unexpected behaviors

[christophe-calmejane]

The controller always expect 0 as first unsol for Milan device (as the spec says). The issue is that if the controller restarts (quickly) without properly deregistering first, the device cannot detect the controller restarted (even though it's sending a new REGISTER_UNSOL, because the Milan spec doesn't specify the counter should, or not, reset), and will continue sending unsol using its current sequence.