That OVERLAPPED fix seems to have made flow control useful as well

(presumably Windows's serial buffer is actually _filling up_,
causing an XOFF to be sent, now that my dodgy I/O code isn't causing
it to leak). So I think I'll switch the default flow control to
XON/XOFF, since it actually seems to do something now.

[originally from svn r6829]

settings.c

@@ -768,8 +768,8 @@ void load_open_settings(void *sesskey, int do_host, Config *cfg)
     gppi(sesskey, "SerialSpeed", 9600, &cfg->serspeed);
     gppi(sesskey, "SerialDataBits", 8, &cfg->serdatabits);
     gppi(sesskey, "SerialStopHalfbits", 2, &cfg->serstopbits);
-    gppi(sesskey, "SerialParity", 0, &cfg->serparity);
-    gppi(sesskey, "SerialFlowControl", 0, &cfg->serflow);
+    gppi(sesskey, "SerialParity", SER_PAR_NONE, &cfg->serparity);
+    gppi(sesskey, "SerialFlowControl", SER_FLOW_XONXOFF, &cfg->serflow);
 }
 
 void do_defaults(char *session, Config * cfg)