Consider inf/nan when validating float ranges.

[kape1395]

In some cases, the min/max boundaries for floats are :infn and :inf. These cases have to be considered.
Not sure if such validation conforms to the standard.
Also, I don't know if the value itself can be inf/infn/nan. In such a case, the validation has to be adjusted further.

[gitneko]

Thanks for the PR!

I have never seen so far NaN/Inf+/Inf- for REAL properties in the wild so far, but I haven't worked with every possible BACnet device yet. :)

The BACnet specification simply refers to IEEE single and double precision floating points, so non-numbers are definitely possible for all REAL properties (such as the value itself). There's no explicit statement against it. So that case would need to be handled aswell.

For me the following are missing for the new function:

• Docs
• Typespec
• Tests (and "integration tests" for the analog objects that the min/max validation works properly)
• And since there is more than one clause one function head for "pretty" variables in the docs (def float_validator_fun(value, object) works just fine)

Once you tackle these points, I'll check it out locally and if everything is OK, I will merge it.

Once again, thanks for your work and the time you put into using and improving this project.