Vega Strike Code

This is an attempt to reset fairness in scheduling in the game. Basically, since we're not threaded we have to decide how many physics updates we want to run per draw and this more closely tunes things to the settings the player makes with Computer (simulation_atom size). It's important when testing this commit to see if it should be reverted to make sure your settings are matched to your computer. Slow framerate will == jerkier behavior, but game speed should remain divorced from framerate for the most part

-O options are now their own flag and independent of Build Mode or CPU opts. You can mix and match all 3 settings. We still need to do a better job at ensuring flags work with the current gcc version in use, however.

removing a Unit pointer from a collection should only need to traverse until the first unit match, we dont put multiple copies in a single list. Also some minor const optimizations and unit_collide was re-creating a function already in collection

Not sure why one would think gravitational units aren't on the drawlist, they even have a special case where they are drawn regardless of distance

in_joystick had a condition that gcc would sometimes optimize out. Basically, numaxes should equal the smaller of what SDL is reporting or MAX_AXES... original code didn't adhere to this limit, though luckily MAX_AXES is so big we never encountered issues. Fixed it to behave logicall.

some minor cleanups (shouldn't do any harm

Revert last "optimization" - misguided it was, because -O3 was missing

Avoid copying arguments in XVector

Avoid numeric imprecision in lerp

Optimize particles by killing them faster