RoundOp not align with pytorch result

[wyg1997]

Summary

Pytorch kernel implements the “round half to even” to break ties when a number is equidistant from two integers (e.g. round(2.5) is 2). （ https://pytorch.org/docs/stable/generated/torch.round.html#torch.round ）

Code to reproduce bug

import oneflow as flow
import torch
import numpy as np

arr = [0.5, ]

of_in = flow.tensor(arr)
torch_in = torch.tensor(arr)

of_out = flow.round(of_in)
torch_out = torch.round(torch_in)

if not np.allclose(of_out.numpy(), torch_out.numpy()):
    print("Not Equal")
    print("of_out: ", of_out.numpy())         # [1. ]
    print("torch_out: ", torch_out.numpy())   # [0. ]