Don't kill the managed proxy if a transport failed to launch

We currently do:

  } else if (!strcmpstart(line, PROTO_SMETHOD_ERROR)) {
    if (mp->conf_state != PT_PROTO_ACCEPTING_METHODS)
      goto err;

    parse_server_method_error(line);
    goto err;

which means that we destroy managed proxies that were supposed to spawn multiple transports but failed to launch one of them.

I think we should log a warning, but keep the managed proxy around. Some transports are better than no transports; and the warning message should hopefully prompt the user into fixing the issue.