Fix unnecessary rig re-open cycles by refactoring rigctl_parse return values

[mikaelnousiainen]

The current command loop in both rigctl and the rigctld server perform rig close/open cycles in case of hardware or I/O errors. However, the logic to detect these errors is not very accurate, as rigctl_parse returns value 2 (indicating re-open is required) for all RIG_E* errors from rig_ commands. There are many errors, e.g. "invalid parameter", which do not indicate an issue in the rig communication and re-opening the rig is totally unnecessary in this case.

Re-opening the rig causes real trouble when there are multiple clients connected to rigctld server and one of the clients issues a command resulting in "invalid parameter" error, for example. This will lead to the rig being unavailable for some time all other clients. See #612 for a more detailed explanation.

This PR includes:

1. New, more deterministic return values in rigctl_prase function. The function returns the negative RIG_E* errors as they are, uses RIG_OK for indicating a successful command and two positive integers RIGCTL_PARSE_END (1) and RIGCTL_PARSE_ERROR (2) to indicate end-of-stream and other internal parsing errors -- both of which will make the command loop exit.

2. New macro RIG_IS_SOFT_ERRCODE for determining if a particular error code is a "soft error", indicating it is not a hardware or I/O fault and cannot be potentially fixed by re-opening the rig. The soft errcodes are: RIG_EINVAL, RIG_ENIMPL, RIG_ERJCTED, RIG_ETRUNC, RIG_ENAVAIL, RIG_ENTARGET, RIG_EVFO and RIG_EDOM.

3. Some fixes to Icom backend to remove ACK/NAK detection in commands that get/read data from the rig, as the ACK/NAK bytes are present in responses to commands that set/change the rig state only.

[mikaelnousiainen]

That was fast! Thanks :)