Gracefully handle portal disconnections lasting more than a few seconds about firezone HOT 4 CLOSED

The intention for this issue is to have a reliable reconnect / connection down detector than we can use to issue callbacks to the client on the state of the tunnel/adapter and not the other way around. In essence, doing our own path monitoring.

This cross-platform approach would make it easier to communicate to the user the real connection status of their Firezone client and not rely on a mixture of OS-specific path monitoring and connlib errors to derive a client state to show to the user. That seems like an endless rabbithole of UI state bugs

from firezone.

connlib should be able to handle network partitions lasting up to X seconds (configurable) global timeout with the portal, after which point it calls onDisconnect without an error (so that the token is not cleared). X seconds is expected to be on the order of ~300 seconds (5 minutes), but could be longer.

How will it be configurable? will it be handed down by Session::connect?

from firezone.

@conectado Could just be a constant/hidden env var for now that we can override for tests.

from firezone.

@jamilbk kk, I will use an env var and default to 5 mins if it's not there

from firezone.