Vega Strike Code

make erase take one argument. Should be the last of the optimizations and cleanups for unitcollection. It's _really_ fast now... But AI is still horribly innefficient, traversing subunit lists of its parents for each subunit (that's bad and just feels wrong).

It looked like for some reason this while loop got mangled

Revert most of my change made with the commit for modifying simulation_atom...doesn't seem to have the desired effect

Further optimization of unit collection and iterator classes. I moved active iterator to a fixed length array slot based system. We currently never peak over 3 concurrrent iterators on a given list so i set it to 5. By having a fixed length array, we drastically improve speed of looking up other active iterator positions and thus we're able to avoid keeping lists of null holder positions. a unit collection should only ever return NULL when at the end of the list. I also optimized = operator

Add more py2 compatibility imports, not sure they're *required*, but
they're correct.

Merge the rest of py3 branch. Also make it compatible with py2, for the
lulz (and because it might suit mods)

Un-revert r13505 & r13507, prior to merging py3 python changes

Forgot to commit these images

Rename int32bit and char8bit to uint32bit and uchar32bit, to reflect
their unsignedness

only include ftree parallel loops option if we want > 1 ... not sure if gcc checks for 1 to not include unecessary gomp in such a situation, so include it conditionally. Makes sense