Normalization Term Impact on Gradient and Hessian Scaling

[RektPunk]

It was noted that the normalization term is affecting both the gradient and Hessian, while other aspects remain unchanged. Currently, for the normalization parameter to take effect, it needs to be set to a significantly large value.
However, if we divide by N (the number of data points), the normalization effect should be noticeable with smaller values.

For future updates, simply divide the gradient and Hessian by the number of data points to address this issue.

MQBoost/mqboost/objective.py

Lines 85 to 87 in 7e05273

	_err_for_alpha = _y_train[alpha_inx] - _y_pred[alpha_inx]
	_grad = grad_fn(error=_err_for_alpha, alpha=alphas[alpha_inx], **kwargs)
	_hess = hess_fn(error=_err_for_alpha, alpha=alphas[alpha_inx], **kwargs)