[rcore] Matrix determinant calculation speed up (enhancement)

[raylibfan]

More then 2.0x faster Matrix determinant calculation can be achieved

In raymath.h line 1467 we have such realisation:

// Compute matrix determinant
RMAPI float MatrixDeterminant(Matrix mat)
{
	float result = 0.0f;

	// Cache the matrix values (speed optimization)
	float a00 = mat.m0, a01 = mat.m1, a02 = mat.m2, a03 = mat.m3;
	float a10 = mat.m4, a11 = mat.m5, a12 = mat.m6, a13 = mat.m7;
	float a20 = mat.m8, a21 = mat.m9, a22 = mat.m10, a23 = mat.m11;
	float a30 = mat.m12, a31 = mat.m13, a32 = mat.m14, a33 = mat.m15;

	result = a30 * a21 * a12 * a03 - a20 * a31 * a12 * a03 - a30 * a11 * a22 * a03 + a10 * a31 * a22 * a03 +
		 a20 * a11 * a32 * a03 - a10 * a21 * a32 * a03 - a30 * a21 * a02 * a13 + a20 * a31 * a02 * a13 +
		 a30 * a01 * a22 * a13 - a00 * a31 * a22 * a13 - a20 * a01 * a32 * a13 + a00 * a21 * a32 * a13 +
		 a30 * a11 * a02 * a23 - a10 * a31 * a02 * a23 - a30 * a01 * a12 * a23 + a00 * a31 * a12 * a23 +
		 a10 * a01 * a32 * a23 - a00 * a11 * a32 * a23 - a20 * a11 * a02 * a33 + a10 * a21 * a02 * a33 +
		 a20 * a01 * a12 * a33 - a00 * a21 * a12 * a33 - a10 * a01 * a22 * a33 + a00 * a11 * a22 * a33;

	return result;
}

It takes 72 multiplication to calculate 4x4 matrix determinant as one can see.

Meanwhile, accordinly to this Wikipedia's article https://en.wikipedia.org/wiki/Laplace_expansion , this calculation can be reduced to (4 + 4 * (3 + 3 * 2)) = 40 multiplication by decreasing matrix size from 4x4 to 2x2 using minors.

Such algorithm is used in blender 3d application source code as for example: https://github.com/blender/blender/blob/main/source/blender/blenlib/intern/math_matrix_c.cc from LINE 1837

So we can replace those existing code fragment from above by something like this (using raylib function naming convention):

RMAPI float DeterminantM2(const float a, const float b, const float c, const float d)
{
	return (a * d - b * c);
}

RMAPI float DeterminantM3(float a1, float a2, float a3, float b1, float b2, float b3, float c1, float c2, float c3)
{
	return (a1 * DeterminantM2(b2, b3, c2, c3) - b1 * DeterminantM2(a2, a3, c2, c3) + c1 * DeterminantM2(a2, a3, b2, b3));
}

// Compute matrix determinant
RMAPI float MatrixDeterminant(Matrix m)
{
	return (m.m0  *  DeterminantM3(m.m5, m.m6, m.m7, m.m9, m.m10, m.m11, m.m13, m.m14, m.m15) -
		m.m4  *  DeterminantM3(m.m1, m.m2, m.m3, m.m9, m.m10, m.m11, m.m13, m.m14, m.m15) +
		m.m8  *  DeterminantM3(m.m1, m.m2, m.m3, m.m5, m.m6,  m.m7,  m.m13, m.m14, m.m15) -
		m.m12 *  DeterminantM3(m.m1, m.m2, m.m3, m.m5, m.m6,  m.m7,  m.m9,  m.m10, m.m11));
}

I wrote some simple test application to check perfomance boost. New or "blender style" function is more then twice faster, then current "raylib style".

Issue Screenshot

P.S. I am crab-handed and can not provide a PR :)