Added army instancing and a CPU cost display

I changed infantry rendering to use instances and share pose work, then reduced repeated neighbor and fire-collision searches.

F8 now exposes the CPU costs of simulation and rendering stages alongside triangle and draw-call counts. The display explicitly leaves GPU timing unmeasured.

“I changed infantry rendering to use instances and share pose work, then reduced repeated neighbor and fire-collision searches.”

The recording shows the counter during a native battle. These changes are still being tuned, and I am not claiming a particular frame rate from this short capture.