Updated testeyes.c.  It now shows some kind of gibberish Missingno-type picture somewhere on the screen and, possibly, hangs the window system.  On the upside, the Missingno is definitely a non-rectangular window.

Put in a couple of fixes that I realized hadn't gotten done when I wrote out the SCRUM stuff in TODO.  Added SCRUM listings in TODO.

Ported over, to the best of my ability, the code for Win32 shaped windows and patched in the correct C syntax and coding conventions of SDL.

Updated test code, updated win32 code a bit (still not complete, but hopefully tonight), and removed the last vestiges of ellipse and polygon drawing support.

Added SDL_x11shape.h and removed unnecessary header includes from SDL_x11shape.c.  This should help Andreas and Sam to compile the new code.

Replaced the model of treating a window's shape as a render target with treating a window's shape as a surface passed into an SDL_SetWindowShape() function.  I'll send this off to Andreas and Sam and start coding.

Stubbed out the new get/set parameter functions, added the start of the enum, union, and struct for those parameters themselves, and added doxygen comments throughout the SDL_shape.h header.

Moved the SDL_Eyes test program for shaped windows into the Hg repository and got it building and linking as a test.

Added the standard headers and footers that make SDL_shape.h and SDL_shape.c integrate properly into the build.

Changed API for shaped windows to make it much, much simpler.  A shaped window's shape mask is now accessed as a render target, which means we can get at it with blit and draw functions, as well as (most likely) OpenGL.

I'm actually going to have to alter this proposed API a fair bit, but it's coming along.  I want to keep it as orthogonal to the rest of SDL as possible.