Merge threadbase.h into thread.h since we no longer need to have “secret”
thread implementations; all the platforms we care about have std::thread
or just don’t support threads at all.

Since C++11 threads are available on every platform that has threads, we
no longer need our pthread implementation. The LOL_FEATURE_CXX11_THREADS
macro is gone, too, and we now just use LOL_FEATURE_THREADS.

Note that it is still necessary to link with -lpthread or -pthread on
some platforms, so we don’t throw that part away.

Add mingw-std-threads project to external/ for full C++11 thread support, fix
an incorrect Win32 fix I did in PEGTL, work around mingw’s <io.h> declaring a
conflicting eof symbol, and fix a capitalisation error in log.cpp.

The C++ standard (21.5 Numeric conversions) says that %lld must be available,
however mingw still does not support it and in the meantime the most practical
solution seems to use long int instead of long long int.

It merely performs an std::swap() on both array elements, but it’s
apparently convenient to have.

Using “Array” instead of “array” is now fully deprecated, and I replaced
the Log::Debug() etc. functions with msg::debug() etc. because log:: was
unavailable due to being a maths function.

Added up-to-date copyright notice and BOM to modified files, just in case.

The containers no longer force the user to use the ptrdiff_t type for
size information. For convenience, size() now always returns an int,
and the size_s() method offers a way to handle arrays with more than
2 billion elements. Internally, we still use ptrdiff_t, though.

Since so much code had to be changed, I took the opportunity to get
rid of capitalised accessors for classes that are lowercase.

Though it still is not possible to create a new window from a thread that did not init SDL

We now have a box_t<T,N> template class. Which means we can have
4-dimensional AABBs. Why? Because we can.

In release mode our asserts are ignored for performance reasons and all
the code they may be executing is disabled, so no side effects will happen.
We can’t put programming logic in asserts.

Scene rendering logic is now better

The “f” suffix is valid GLSL, but WebGL does not like it so let’s disable it.

Box objects (AABB) now have the long awaited extent() and center()
methods. Corners are now called “aa” and “bb” for convenience. The
WorldEntity class uses a box3 for its AABB instead of two vec3s.
This allows us to simplify a lot of verbose code.

Disallow firing when no ammo is active, tweak laser position, get rid of
the old tileset, add a title screen.

My Ludum Dare 32 entry is now in the main games/ directory and is
called “Size-o-Tron 3000” because… because, that’s all.

When a thread is destroyed during the atexit() callback cleanup, it cannot
be joined and sits there doing nothing if the CRT is Visual Studio 2012 or
2013. Microsoft has stated that they would not be backporting the fix from
Visual Studio 2015. As a workaround we call ExitThread() forcibly after
the thread exits.

Text objects now support character spacing (as a fraction of the character
width) and text scaling. Also, the alignment enum is now safe.

For some reasons std::thread can’t be joined in Visual Studio 2013. Needs
more investigation but right now it’s Ludum Dare and I don’t have time.

This will be a 2D puzzle platformer. The theme this year is
“Unconventional Weapon”.

This sort of pImpl inheritance code code was weird, I got rid of it
and added a new pointer in the TileSet class instead.

Lua only supports the "double" type for now, so storing a 64-bit integer
would lose information unless it's encapsulated in a more complex type.
I prefer to disable these accessors for now.

Reworked Primitive logic for scene rendering
FileUpdateTester works now .....