Add logistic regression multi classification ovr

[magic-hya]

Pull Request

What problem does this PR solve?

增加多分类ovr功能

Issue Number: Fixed #252

Possible side effects?

• Performance:

• Backward compatibility:

[magic-hya]

目前遇到问题：
使用
roc_auc_score(y.values, result2[0], multi_class='ovr')
函数时，该函数需要校验prob的求和和1的误差，
误差校验函数：

# validation of the input y_score
if not np.allclose(1, y_score.sum(axis=1)):
    raise ValueError(
        "Target scores need to be probabilities for multiclass "
        "roc_auc, i.e. they should sum up to 1.0 over classes"
    )

其中校验函数的默认参数为 rtol=1.e-5
spu中计算prob代码为：
prob = preds / preds.sum(axis=1).reshape((preds.shape[0], -1))
求和后，结果如下，误差精度超过1.e-5

[0.99998796 0.99999934 0.99998623 0.999986   0.99999976 0.9999895
 0.9999957  0.9999958  0.99999976 0.9999894  0.99998903 1.
 0.99999917 0.9999858  0.9999884  0.999993   0.9999982  0.9999987
 0.9999858  0.99999857 0.9999856  1.         0.9999879  1.
 1.         1.         1.         0.99999666 0.999994   0.99999887
 0.9999904  0.9999883  1.         0.99999344 0.9999994  0.99999917
 0.9999925  0.9999993  0.9999993  0.99999154 0.99998975 0.99999785
 0.9999871  0.9999868  0.9999989  0.99999994 0.9999863  0.99998987
 0.9999995  0.9999865  0.99998564 0.9999861  0.9999868  0.99999475
 0.9999995  0.9999935  0.99999225 0.9999963  0.9999888  1.
 0.9999999  1.         1.         1.         0.99999446 0.99999964
 1.         0.9999996  1.         0.9999998  1.         0.99999964
 1.         0.99998564 1.         0.9999913  0.999997   0.9999975
 1.         0.9999993  0.99999994 0.99999964 0.99999994 1.
 0.9999845  0.9999995  0.99999344 0.9999964  0.99999905 1.
 0.99999785 0.9999969  0.99999994 1.         1.         0.99998987
 1.         0.99999994 1.         0.9999984  1.         0.99999994
 0.9999529  0.99999917 0.99999964 0.9999957  1.         1.
 1.         0.99999994 0.99999875 0.9999993  0.99999523 0.9999989
 1.         0.9999908  1.         0.9999999  1.         0.99999934
 1.         0.99999213 0.99999803 1.         0.99999994 1.
 0.99996746 0.9999928  1.         0.9999999  0.99998355 0.99999964
 0.9999997  0.99999976 0.99999976 0.99999994 1.         1.
 0.9999999  1.         0.99999964 0.9999957  0.9999997  0.99999994
 1.         0.9999999  0.99999964 0.9999999  0.9999997  1.
 0.9999998  0.99999976 0.9999969  0.99999994 0.9999999  1.
 0.9999999  1.         0.9999996  0.99999726 0.99999994 0.99999976
 1.         0.9999998  0.9999996  0.99999994 1.         0.9999999
 0.9999997  1.         1.         0.99999976 0.99999815 1.
 1.         1.         1.         1.        ]

[deadlywing]

@magic-hya Hi，抱歉前段时间比较忙，review不及时，一些意见供参考，欢迎讨论～

[deadlywing]

LGTM