Support timeouts on Request/Response protocol level

[xgreenx]

Problem description

Any error(it includes timeouts) that occurs during the request-response protocol is ignored. If we receive an error, the parts of the blockchain that await the response will be paralyzed. In the case of the fuel-core-sync, it may stuck the synchronization. In other cases, it can consume resources forever.

Solution

Instead of silently removing the channel related to request_id we need to send either an error or None. It allows other parts of the project to handle it and request it again(maybe from another peer).

Implementation details

The approach with error requires modification of the response type. But it triggers cascade changes in the places where we started a request. It will allow us to check all related places and to be sure that we handled an error response correctly and at least created retry logic(or any other logic that suits us to be sure that the node is in the "fine" state after this request).

The main goal of this issue is to have timeouts for requests. We need to add an integration test that verifies that. Also, maybe we need to make a timeout threshold configurable(the default value could be 20 seconds).