Fix frequency input logic

[zcsahok]

Changing frequency by entering it in the call input field is a handy feature, but it has some rough edges.

1. if an out-of-band frequency (all digits) is entered then it is treated erroneously as a callsign
   

2. similarly, when there is no rig control then all numeric values are taken for a callsign
   

Reworked sendqrg() to use plain_number() to check if input is a number. Originally atof() was used to parse it, presumably to support fractional kHz values. As dot is used to configure bandmap this part of the feature was broken anyway, so I replaced it with a simple atoi().
In logit.c moved the clearing in case of freq up to the check itself and thus reduced the indenting of the code actually doing relevant stuff.

[dl1jbe]

Looks good from my side.

While looking up the changes I found an error in callinput.c line 1257. send_bandswitch() expects an frequency as parameter but gets an bandindex. Maybe you can add a fix for it too.

[zcsahok]

Fixed the argument for bandswitch. Probably nobody uses the parallel(!) port info cwdaemon can provide, so this went unnoticed.

Also fixed plain_number(): it returned true for an empty string. When called from callinput.c it is always ensured that the input is not empty, that's why this wasn't a problem until now.
The effect of treating empty string as a number broke keyer window if the input field was empty.

[dl1jbe]

Fixed the argument for bandswitch. Probably nobody uses the parallel(!) port info cwdaemon can provide, so this went unnoticed.

Thanks.

Also fixed plain_number(): it returned true for an empty string. When called from callinput.c it is always ensured that the input is not empty, that's why this wasn't a problem until now. The effect of treating empty string as a number broke keyer window if the input field was empty.

Fine too.

Whole PR looks good. Should be merged in.