Experimental software optimizations

Contains a few experimental software optimizations, specifically for 64-bit systems. Optimizations currently applied:

• Texture offsets calculated with multiplication rather than division.
• Draw function variables are now pointer-sized, which avoids integer promotion within performance-critical loops.
• Software wrapping fix is removed on 64-bit systems, as they are no longer necessary thanks to large integer sizes.
• Backbuffer is transposed, allowing the CPU to take advantage of write cache more efficiently. Reverted due to to many problems.
• Buffer flipping is now done directly to video memory rather than via SDL - this also have the side effect that the game runs in actual screen resolutions, which prohibits certain resolutions from working in fullscreen if the monitor doesn't support it.