[tz] proposed changes for Win32 and a improved mktime() algorithm

[Kees Dekker]

>The general rule is that if the platform isn't supported by its supplier any 
>more then we don't need to support it either. HP says they will support HP-UX on 
>Itanium until at least 2025, so it's still viable. However, some old HP-UX 
>compilers are not supported by HP any more, so just saying "I need this for 
>HP-UX" is not enough: the old unbundled C compiler is long obsolete, for 
>example, and the PA-RISC platform is obsolete too.

HP PA-RISC is a different story, but also for HP UX Itanium, the number of fixes + conformance to e.g. C++11 (or later) is quite low.
We are using aCC 06.28.02 (=latest one AFAIK), but it is not updated since past 1.5 year. I guess, but can't prove, that 
HP will limit releasing patches to a very low level, long before 2025 is reached. Their 2025 boundary is also a little bit theoretical,
as the most recent Itanium hardware was launched 5+ years ago, and HW has its own life-cycle policy. The 2025 date is related
to the OS, but is useless if there is no supported hardware anymore (see also: https://www.hpe.com/h20195/V2/getpdf.aspx/4AA4-7673ENW.pdf?ver=Rev.%205).
The end of support date for HPUX11.31 is actually 31-Dec-2020 (IA64). As PA-RISC is not sold for years, the 2025 date is imaginary, because
none of the PA-RISC hardware is still supported.

>Microsoft is another animal: they don't care about whether native MS-Windows 
>conforms to POSIX, and we don't have the resources to maintain a port to their 
>idiosyncratic system. Luckily, we don't need to, as POSIX-like environments such 
>as Cygwin and WSL are readily available for MS-Windows, and though not perfect 
>they're good enough.

I know, but Cygwin is no option for us. As you could see with my patch, I got all working :-).