[META] Improved connection loss handling

[fredo]

This is not an easy problem as it can open vulnerabilities easily. If connection is broken to any external RPC then we must stop the state machine.
The team decided to handle it with better notifications for now to give a better user feedback.

Soft measurements:

• Notification in case of network connection loss #7212
• Add retries to the http requests #7211
• Restart raiden automatically in case of an exception #7213

[ulope]

One more option we discussed:

• Add a soft-restart feature that would allow the client to self restart from the outer layer (cli) in case of an exception, removing the need for the user to setup a process manager