fix: scanner.next() won't automatically recover when server side not return ERR_OK

[foreverneverer]

if next() rpc not response ERR_OK, it will mark _encounterError=true, and all the next() loop will directly return error. so if server side occur transient ERR_TIMEOUT,ERR_SESSION_RESET,ERR_OBJECT_NOT_FOUND,ERR_INVALID_STATE problem, it won't automatically recover after the server side recover. This pr resolve the fix when encounter the rpc failed.

Manual Test

• start pegasus server onebox and scan task
• stop one replica instance, and then client report ERR_SESSION_RESET and update config automatically
• client send next rpc and receive CONTEXT_ID_NOT_EXIST for using new config, which will triger asyncStartScan with current hashKey and sortKey as start index
• client recover and receive right value
• re-start instance that has been stopped and set_meta_level=true to rebalance the onebox cluster
• client receive ERR_INVALID_STATE and update config automatically again
• client send next rpc and receive CONTEXT_ID_NOT_EXIST for using new config, which will triger asyncStartScan with current hashKey and sortKey as start index
• client recover and receive right value