Fix sphere vertex sampling: compare squared radius to squared bound

In-sphere rejection sampling compared x*x+y*y+z*z (a squared magnitude)
against the un-squared max_radius, so the accepted region had radius
sqrt(max_radius) instead of max_radius. Square the bound. The sampling
cube already spans [-max_radius, max_radius] per component, so rejection
sampling stays valid.

Fixes #109

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Author: zhaozhiwen

gemc/gparticle/gparticle.cc

@@ -166,7 +166,7 @@ G4ThreeVector Gparticle::calculateVertex() {
 				z      = randomizeNumberFromSigmaWithModel(0, max_radius, gutilities::uniform);
 				radius = x * x + y * y + z * z;
 			}
-			while (radius > max_radius);
+			while (radius > max_radius * max_radius);
 
 			// Offset the sampled point by the nominal vertex.
 			x = x + v.x();