Sign and type issues with XNET status codes

[d-bohls]

Pulled from discussion in #233.

There are a couple of issues with status codes:

Sign issue:
The C API treats status as an i32. Positive codes are a warning and negative codes are an error. The python API treats status as a u32. It's an error if the MSB is set, and a warning otherwise. If the user is dealing with the status code in hex, this is no big deal. However, equivalent error codes as int will not match between the C API and the python API. This doesn't seem ideal to me. Do we care?

Type issue:
In _cconsts.py line 8 we define NX_STATUS_ERROR = (0x80000000) which comes from the C API and is used as a bit mask for the MSB on i32 status codes. When we do this in python 2.7, NX_ERROR_BASE get promoted from an int to a long because 3220582400 (the unsigned value of 0x80000000) does not fit in 4 bytes for a signed python int. If we set NX_STATUS_ERROR to -2147483648 instead (the i32 value of 0x80000000), this promotion to long doesn't happen. Python 3.6 did away with long so NX_STATUS_ERROR stays an int. So far, the only effect I know of due to this issue is that I ran into difficulty trying to add typing checks on status codes, since codes are a long in 2.7 and an int in 3.6, and 3.6 doesn't know about longs.