Fix: OptimizedSpatialIndex quadtree slower than DefaultSpatialIndex

[YJack0000]

Summary

• Fix critical quadtree bugs that made OptimizedSpatialIndex slower than DefaultSpatialIndex (linear scan)
• Implement proper spatial pruning via circle-AABB intersection in _query
• Fix inBounds rejecting valid queries where center is outside node but range overlaps
• Optimize update() to skip remove+insert when object stays in same leaf
• Replace shared_ptr<SpatialObject<T>> with value storage for cache-friendly access
• Remove leftover debug printf in clear()
• Re-enable previously disabled test that now passes with the fix
• Add benchmark test suite for Default vs Optimized comparison

Benchmark Results

Before (buggy)

Scenario	Default	Optimized	Speedup
SimulateFrame 1000 obj, 10 frames	118.68 ms	267.87 ms	0.44x (slower)
Query 5000 obj	156.46 ms	3.10 ms	50.47x
Update 5000 obj	132.67 ms	633.95 ms	0.21x

After (fixed)

Scenario	Default	Optimized	Speedup
SimulateFrame 1000 obj, 10 frames	116.93 ms	15.22 ms	7.68x
Query 5000 obj	156.55 ms	6.23 ms	25.11x
Update 5000 obj	131.25 ms	25.88 ms	5.07x

Test plan

• All 14 existing unit tests pass (including re-enabled QueryFarAwayButWithinRangeReturnsResults)
• Benchmark confirms Optimized is now faster than Default in simulation scenarios
• Verified buggy version is slower via before/after benchmark comparison

🤖 Generated with Claude Code