Codechange: Store text run positions in vector of points.

This simplifies the interlaced vector of x/y positions.

Codechange: automatic adding of _t to (u)int types, and WChar to char32_t

for i in `find src -type f|grep -v 3rdparty/fmt|grep -v 3rdparty/catch2|grep -v 3rdparty/opengl|grep -v stdafx.h`; do sed 's/uint16& /uint16 \&/g;s/int8\([ >*),;[]\)/int8_t\1/g;s/int16\([ >*),;[]\)/int16_t\1/g;s/int32\([ >*),;[]\)/int32_t\1/g;s/int64\([ >*),;[]\)/int64_t\1/g;s/ uint32(/ uint32_t(/g;s/_uint8_t/_uint8/;s/Uint8_t/Uint8/;s/ft_int64_t/ft_int64/g;s/uint64$/uint64_t/;s/WChar/char32_t/g;s/char32_t char32_t/char32_t WChar/' -i $i; done

Change: include fmt.h C++ headers in stdafx.h

This to prevent compilation issues between runs with and without precompiled
headers. Also remove the headers from the rest of the code base as they are
not needed there anymore, although they do relatively little harm.

Codechange: remove #ifdef from .cpp files to exclude features

With CMake, these files are simply not compiled to start with.

Codechange: Store text layout runs directly as values in a std::vector instead of heap allocated.

This reduces memory allocations and heap fragmentation.

Change: [Win32] Use Uniscribe instead of ICU for text caret handling.

This removes the need for the ICU lib on Windows.

Add: [Win32] Text layout using the native Windows Uniscribe library.

Uniscribe is sometimes producing different results compared to ICU, especially
when RTL and LTR content is mixed. Comparing the results to other programs
(like editors or web browsers) leads me to believe that the result are at least
not worse than ICU and possibly better.