This code was mostly written as an experiment, to see just how much
could be done using only the C preprocessor (give or take). Turns out
you can do quite a lot.

In fact, *all* of the tables in this library are generated at compile-time.
The polynomial is #define'd in crc32i.h, and all of the tables are generated
through a combination of enums and preprocessor trickery. Note that this
may cause this code to compile slowly on some machines or compilers.

At the moment it is hardcoded for x86-64 and gcc, but it could be adapted to
other compilers if they also have features like e.g. alignas() or whatever.
You'd also need to disable the x86 code which isn't difficult but is a bit
annoying.