Code organisation

[Markus Mützel]

On Windows, the functions "getenv" and "putenv" still fail with Octave dev if
one wants to query or set environment variables containing non-ASCII characters.
This implies that some functions (e.g. "which") are failing if a user with
non-ASCII characters in their name tries to run Octave.

I have a patch that switches to using the Unicode API for these functions.
However, I'm unsure how that should integrate into the code base.
For "putenv" that's pretty clear because we are already using a wrapper for
gnulib.
"getenv" is in the namespace "octave::sys::env".
Should I add the OS dependent code directly in "octave::sys::env:do_getenv"? Or
is it cleaner to add a new wrapper "octave::sys::getenv_wrapper" in lo-sysdep
(see attached patch)?

Thanks for any advise
Markus

getenv_unicode.patch
Description: Binary data