Task timeouts are not respected while reconnecting in iscsi_tcp_service

[theaugy]

Immediately after checking that iscsi->fd is valid, iscsi_tcp_service will check if there is a pending reconnect:

	if (iscsi->pending_reconnect) {
		if (time(NULL) >= iscsi->next_reconnect) {
			return iscsi_reconnect(iscsi);
		} else {
			if (iscsi->old_iscsi) {
				return 0;
			}
		}
	}

If it finds one, it will either attempt a reconnect and return, or just return (my reading of the code tells me old_iscsi is non-NULL for essentially the entire time we are disconnected).

This is problematic for use cases where the task timeout has been set, as tasks which should be timing out during the disconnect are unable to do so.

I wonder if either moving the call to iscsi_timeout_scan(iscsi); before the pending reconnect block, or adding a call before each return statement, would solve this issue safely.

[theaugy]

I was able to address this issue by:

1. moving iscsi_timeout_scan before the pending_reconnect check, as suggested in the issue
2. having iscsi_timeout_scan "recurse" into old_iscsi