Exclusion routes are not re-evaluated on interface changes

[ekoby]

Exclusion routes (to controllers and edge routers) should be re-created when interfaces change

[qrkourier]

Potentially related issues in #398 and #389

[nirmal0522]

Initially, the device has 2 internet connections up and running via gateways 192.168.6.1 and 192.168.4.1 where 192.168.6.1 is the active (preferred) default route
The tunneler starts and adds an entry to the routing table that effectively pins all traffic going to the public edge router and controller via the currently preferred default route.
34.106.77.247 via 192.168.6.1 dev eth1
174.129.210.139 via 192.168.6.1 dev eth1
A curl fetch to another endpoint hosted service is successful.
The primary internet link through 192.168.6.1 fails and mwan3 on the OpenWrt Linux host switches the preferred default gateway from 192.168.6.1 to 192.168.4.1 (around 75 seconds after the tunneler starts – see timestamps in square brackets).
All subsequent TCP connections from the tunneler end up getting lost because of the pinned routing table entries pointing to a now unreachable gateway (as seen from the logs and the timeout in reaching the service which was previously reachable)