Never restart stopped TCE

[joeyoravec]

Around suspend-to-ram it's likely that SOMEIP-SD and TCP sockets will be impacted by clocks jumping. In case:

• SOMEIP-SD timeout is detected
• clear_client_endpoints clears everything and stops the TCE endpoint
• new OFFER is received

then there's a race condition where:

• add_routing_info ends up creating a new endpoint
• old (stopped) endpoint calls restart which clears sending_blocked_
• new endpoint calls start

and may result in two connected endpoints, depending on execution order. In order to avoid this, never restart a stopped endpoint.