Signed-off-by: zhoufeng <zhoufeng54@huawei.com>

From: @lianliguang
Reviewed-by: @chujinjin
Signed-off-by: @chujinjin

Merge pull request !6061 from kisnwang/fix-fork-thread-lock

Merge pull request !6024 from 张清华/master2

Remove reentrant function in signal handler;
Just close self in client when timeout during transaction;

Merge pull request !5055 from lirongzhen1/int64

Merge pull request !5161 from ZhangQinghua/master

When run multiple processes by 'mpirun',
parent process never quit even Exception happens,
which caused by MPI_Finalize() never returned.

Merge pull request !4392 from xychow/use-float16-in-arm-neon

Merge pull request !4561 from ZhangQinghua/master1

- It provides a unified style to express graph and kernel for user.
- It provides a unified IR to represent graph and kernel for developer.
- It breaks the boundary between graph and kernel.
- It provides more opportunities to do compile optimization.