[rmodels] Removing powf function from CheckCollisionBoxSphere

[maiconpintoabreu]

Removing powf with a similar way to calculate used in CheckCollisionSpheres

Test:

#import "raylib.h"

int main(void) 
{
    // Define a 2x2x2 box centered at the origin
    BoundingBox box = {
        .min = {-1.0f, -1.0f, -1.0f},
        .max = { 1.0f,  1.0f,  1.0f}
    };

    // Test Case 1: Sphere inside the box
    Vector3 center1 = {0.0f, 0.0f, 0.0f};
    float radius1 = 0.5f;
    
    // Test Case 2: Sphere touching the edge
    Vector3 center2 = {2.0f, 0.0f, 0.0f};
    float radius2 = 1.0f; // Distance is exactly 1.0 from box.max.x

    // Test Case 3: Sphere far away
    Vector3 center3 = {5.0f, 5.0f, 5.0f};
    float radius3 = 1.0f;

    TraceLog(LOG_INFO, "--- Collision Tests ---");
    TraceLog(LOG_INFO, "Test 1 (Inside): %s", CheckCollisionBoxSphere(box, center1, radius1) ? "COLLISION" : "NO COLLISION");
    TraceLog(LOG_INFO, "Test 2 (Touching): %s", CheckCollisionBoxSphere(box, center2, radius2) ? "COLLISION" : "NO COLLISION");
    TraceLog(LOG_INFO, "Test 3 (Far Away): %s", CheckCollisionBoxSphere(box, center3, radius3) ? "COLLISION" : "NO COLLISION");

    return 0;
}

The benchmark on my machine the change did not improve a lot but it should improve on low end devices.

Changed the benchmark tool to use bench.h instead of poop (it was too noise)

1 Call
[==========] Running 2 benchmarks.
[ RUN      ] CollisionSuite.Old_Powf
[       OK ] CollisionSuite.Old_Powf (mean 0.051us, confidence interval +- 0.237557%)
[ RUN      ] CollisionSuite.New_Clamp
[       OK ] CollisionSuite.New_Clamp (mean 0.039us, confidence interval +- 0.582900%)
[==========] 2 benchmarks ran.
[  PASSED  ] 2 benchmarks.

10000 Calls
[==========] Running 2 benchmarks.
[ RUN      ] BatchSuite.Old_Powf
[       OK ] BatchSuite.Old_Powf (mean 148.454us, confidence interval +- 1.816232%)
[ RUN      ] BatchSuite.New_Clamp
[       OK ] BatchSuite.New_Clamp (mean 7.330us, confidence interval +- 2.055348%)
[==========] 2 benchmarks ran.
[  PASSED  ] 2 benchmarks.

[raysan5]

@maiconpintoabreu thanks, I'm merging but it's not a change I specially like, the improvement is minimum and it adds dependency on other functions, that is something I always try to minimize. Main reason to merge is consistency with CheckCollisionSpheres()...

[maiconpintoabreu]

@raysan5 thanks, the main reason of the PR was to make it silimar to the other methods and more readable, the benchmark was to make sure that was no performance regression, the numbers are too small to see the improvement indeed.