redo stops normal in case of tcp_error

[jonasrichard]

Here redo:handle_info redo stops with normal reason in case of tcp error. However in case of tcp_closed if reconnect is false it stops with a different reason. It can cause connection pools like cuesport not to restart children which stopped with a normal reason (transient children). I think it would be better to have some different stop reason and I also implement the reconnect branch of handle_info in case of tcp_error. What do you think?