Replace glib with utfcpp

[kintel]

This is just an experiment to see if replacing glib with utfcpp would result in nicer and still correct code.
Can be killed if it's a dead end.
As discussed in #594

[t-paul]

Hmm, the interface certainly looks a bit nicer, although there's a bit of improvements possible with the glib stuff too, right now it's just sprinkled everywhere where needed (which is not glibs fault).

Anyway, I'll collect my personal pro/cons here :-)

pros

• Small header-only lib with C++ interface
• Included in most distros
• Seems to be actively maintained, small number of changes

cons

• not in MXE (probably easy to do though as header only)
• never really had any issues with glib for the Linux and Windows builds, no idea about Mac OS though
• Distros seem to ship old version (e.g. Ubuntu 18.04/18.10 have 2.3.4 with 2.3.5 released 2016-11-06 / 2.3.6 just now released)
• Using a different library because of "glib is big and I don't like it" seems to be a quite weak argument as it does focus on a not very important fact of the used code
• Surprisingly small adoption, Debian/Testing finds 7 build-depends for libutfcpp-dev where libglib2.0-dev shows 1006

info

• Proposal for Boost.Text at C++Now 2018 (video: part1 / part2)

[kintel]

Thanks for your input. I figured an actual implementation will help us evaluate this better, so let's let it soak for a while :)

[kintel]

It seems like glib is getting more unwieldy for each release. We still manage to keep it somewhat in shape, but it would be nice to be able to rely on something slightly easier to build.