Several, mostly minor, points to solve for the Windows platform

[arjenmarkus]

I have just pushed my changes to support the three Windows platforms:

• Cygwin: no changes required
• MinGW: as I have is no suitable gfortran compiler for this platform, current installed version is 7.3.0, I had to work around several issues. Tested as much as possible.
• plain Windows: using Intel Fortran and MSVC++. The changes have been blended into the original code via preprocessor macros

During the development of this new code, I discovered:

• plain Windows: the makefiles do not lead to a test executable. I must do that manually.
• the test program contains several "unixisms" - symbolic links are in some way supported on Windows, but they seem to work in a different way than on Linux. So the symlink routine always throws an error. This prevents, however, much of the rest of the program to be run.
• it also assumes that the directories /home and /bin exist.
• the function ismount may be troublesome on Windows, that requires further investigation. Specifically: what does it mean for that OS?
• the gfortran compiler on MinGW (7.3.0) has problems with split() but also with relpath() (connected to repeat()).
• the library currently ignores the backslash as a directory separator. While the forward slash is fully (?) supported on Windows, it will be useful to support the backslash as well.
• we may need to handle disks (C:, D:, ...) separately
• what about file names that contain non-basic ASCII characters? They are a nuisance in my opinion, but they are used.

[arjenmarkus]

I should probably split up the list into smaller issues, so that we can work on them in an orderly fashion.