[ab] Fast pixel access

[Luke Crook]

Anthony Fairchild <anthonyf at ...> writes:

> 
> Hey Luke,
> 
> With your changes the rotations are now blazing fast!  

Glad to hear it. How much faster? The generator functions return closures that 
cache a lot of the SDL surface details that were previously being retrieved on 
each pixel read/write. This does away with at least 5 CFFI calls per read-
write. And the surface is no longer locked/unlocked per read/write, which is an 
additional 3 CFFI calls.

- Luke