This adds support to the CMake build files to support building
HarfBuzz-GObject, which is actually a little bit complicated,
as we need to:

-Run glib-mkenums, which is a Python (for newer GLib/GObject) or PERL
 (for older, pre 2.53.4, GLib/GObject) script.  This matters more for
 Visual Studio builds as shebang lines are not supported, so we need to
 test-run it with PERL or Python to determine which is the correct
 interpretor to use.

-Next, we need to replace strings in the sources that we obtain from
 running glib-mkenums.  So, the solution here is to use a small utility
 CMake script for our purposes here, to maintain maximum compatibility.

-Ensure that things do work in the Visual Studio IDE builds.

Include the fallback sources in the build, and update
src/Makefile.sources so that it can be read by the CMake build files.

Fix a typo in the DirectWrite configure option so that we can properly
enable DirectWrite builds.

Also, when building the utility program, install them as well.

The new unorm2_normalize has been public API since ICU 4.4, while
the old unorm_normalize has been deprecated since ICU 56.

Now, beginning with ICU 60, unorm_normalize will also be annotated
U_DEPRECATED and trigger the compiler -Wdeprecated-declarations warning.

c:\projects\harfbuzz\src\hb-set-private.hh(151): error C2327: 'hb_set_t::page_t::v': is not a type name, static, or enumerator [C:\projects\harfbuzz\build\harfbuzz.vcxproj]

error: static_assert expression is not an integral constant expression

Fixes clang "non-const reference cannot bind to vector element" error.

Fixes https://github.com/behdad/harfbuzz/pull/23

Also fix hb_assert_constant_t.

Simplifies compatibility issues by centralizing all boilerplate code
in hb-private.hh.

We might want to tweak this some more.  For now, never skipping over
it is better behavior than always skipping.

Part of https://github.com/behdad/harfbuzz/issues/554

CoreText "point"s are not typographic points, but CSS pixels.  Ie.
they are 96 per inch, not 72 per inch.

Fixes https://github.com/behdad/harfbuzz/issues/435

* Clarified wording of syllable/cluster behavior in Uniscribe in final_reordering; changed one other probable typo.

* Additional syllable/cluster swap in comments for final reordering and for initial-reordering matra decomposition.