It uses heartbeat interval and election timeout to estimate the expected request timeout. This PR helps etcd survive under high roundtrip-time environment, e.g., globally-deployed cluster.