[Gptfdisk-general] Bug fix patch for diskio-windows.cc

[Edward E. <dev...@gm...>]

On Windows, using UINT32_MAX (which is of course 1<<32 - 1) as a scaling
factor to get a uint64_t is causing the sector total to be counted wrong,
affecting disks/images >= 4GiB. Among other things, gdisk wants to "fix"
the secondary GPT by moving it back one sector (plus one sector for each
additional 2TiB). This patch (on current master) uses GetFileSizeEx(),
which returns file size as an int64_t. This API was available possibly as
far back as Windows 98, so switching to it shouldn't be a problem.
Re-compiled on Windows and tested with a known good disk.

I believe this bug caused Chris Murphy's issue in
https://sourceforge.net/p/gptfdisk/discussion/939590/thread/cc21c83b/ ; it
may be involved in
https://sourceforge.net/p/gptfdisk/discussion/939590/thread/6b992837/ as
well.

It may be worthwhile to check if UINT{32,64}_MAX are being used correctly
in all other cases. They look OK to me, but I'm not familiar enough with
the GPT format to be confident of that assessment.