Document/simplify the UpdateMissileVelocity logic

The multiplication by `1 << 21` was unnecessary, with the way the division is performed it doesn't add precision. Missile velocities are 16 bit fixed values so split out the normalisation step in order to make it clearer what the multiplication and use of `v` ends up doing.

Source/missiles.cpp

@@ -107,11 +107,15 @@ void UpdateMissileVelocity(Missile &missile, Point destination, int v)
 	if (missile.position.tile == destination)
 		return;
 
-	double dxp = (destination.x + missile.position.tile.y - missile.position.tile.x - destination.y) * (1 << 21);
-	double dyp = (destination.y + destination.x - missile.position.tile.x - missile.position.tile.y) * (1 << 21);
+	// rotate so that +x leads to the right of screen and +y leads to the bottom (screen space but without the scaling factor)
+	double dxp = (destination.x + missile.position.tile.y - missile.position.tile.x - destination.y);
+	double dyp = (destination.y + destination.x - missile.position.tile.x - missile.position.tile.y);
 	double dr = sqrt(dxp * dxp + dyp * dyp);
-	missile.position.velocity.deltaX = static_cast<int>((dxp * (v << 16)) / dr);
-	missile.position.velocity.deltaY = static_cast<int>((dyp * (v << 15)) / dr);
+	// velocity is stored in screen coordinates so apply the scaling factor to the y axis while normalizing.
+	double normalizedX = dxp / dr;
+	double normalizedY = dyp / dr / 2;
+	missile.position.velocity.deltaX = static_cast<int>(normalizedX * (v << 16));
+	missile.position.velocity.deltaY = static_cast<int>(normalizedY * (v << 16));
 }
 
 /**